Targeting Heat Shock Protein 90 in Pancreatic Cancer Impairs Insulin-like Growth Factor-I Receptor Signaling, Disrupts an Interleukin-6/Signal-Transducer and Activator of Transcription 3/Hypoxia-Inducible Factor-1α Autocrine Loop, and Reduces Orthotopic Tumor Growth

Purpose: Inhibitors of heat-shock protein 90 (Hsp90) may interfere with oncogenic signaling pathways, including Erk, Akt, and hypoxia-inducible factor-1α (HIF-1α). Because insulin-like growth factor-I receptor (IGF-IR) and signal transducer and activator of transcription 3 (STAT3) signaling pathways are implicated in the progression of pancreatic cancer, we hypothesized that blocking Hsp90 with geldanamycin derivates [17-allylamino-geldanamycin (17-AAG), 17-(dimethylaminoethylamino)-17-demethoxygeldanamycin (17-DMAG)] would impair IGF-I– and interleukin-6–mediated signaling and thus reduce pancreatic tumor growth and angiogenesis in vivo.

Experimental Design: Human pancreatic cancer cells (HPAF-II, L3.6pl) were used for experiments. Changes in signaling pathway activation upon Hsp90 blockade were investigated by Western blotting. Effects of Hsp90 inhibition (17-AAG) on vascular endothelial growth factor were determined by ELISA and real-time PCR. Effects of 17-DMAG (25 mg/kg; thrice a week; i.p.) on tumor growth and vascularization were investigated in a s.c. xenograft model and in an orthotopic model of pancreatic cancer.

Results: 17-AAG inhibited IGF-IR signaling by down-regulating IGF-IRβ and directly impairing IGF-IR phosphorylation. Hypoxia- and IL-6–mediated activation of HIF-1α or STAT3/STAT5 were substantially inhibited by 17-AAG. Moreover, a novel IL-6/STAT3/HIF-1α autocrine loop was effectively disrupted by Hsp90 blockade. In vivo, 17-DMAG significantly reduced s.c. tumor growth and diminished STAT3 phosphorylation and IGF-IRβ expression in tumor tissues. In an orthotopic model, pancreatic tumor growth and vascularization were both significantly reduced upon Hsp90 inhibition, as reflected by final tumor weights and CD31 staining, respectively.

Conclusions: Blocking Hsp90 disrupts IGF-I and IL-6–induced proangiogenic signaling cascades by targeting IGF-IR and STAT3 in pancreatic cancer, leading to significant growth-inhibitory effects. Therefore, we suggest that Hsp90 inhibitors could prove to be valuable in the treatment of pancreatic cancer.

Tumor angiogenesis is known to be crucial for growth and metastasis of human pancreatic cancer (1, 2). Over the past decade, multiple factors have been identified that may promote the angiogenic process in pancreatic cancer, which comprise growth factors and cytokines, as well as tissue hypoxia (1, 3–8). Among the growth factors, insulin-like growth factor-I (IGF-I) and its principal receptor IGF-IR seem to play an important role in pancreatic cancer because both ligand and IGF-I receptor itself are overexpressed in human pancreatic cancer specimens (7, 9). Moreover, IGF-IR signaling contributes to invasiveness and survival of cancer cells (6), which in part involves mechanisms that lead to an autocrine activation of IGF-IR in pancreatic cancer (7, 10). In turn, phosphorylation of IGF-IR promotes angiogenesis in tumors by inducing the expression of vascular endothelial growth factor (VEGF; refs. 7, 11, 12). In experimental studies, inhibition of IGF-IR substantially reduced pancreatic cancer growth and angiogenesis (7). Hence, targeting IGF-I/IGF-IR signaling has become an interesting approach for therapy of pancreatic cancer.

Furthermore, the cytokine interleukin-6 (IL-6) and its receptor IL-6R have been implicated in pancreatic cancer growth because elevated IL-6 serum concentrations seemed to be associated with progressive tumor growth and poor prognosis (13). In addition, expression of IL-6R itself has been correlated with increased VEGF-A production in pancreatic cancer (14). Interestingly, IL-6/IL-6R signaling involves the activation of signal transducers and activator of transcription-3 (STAT3), a transcription factor that also plays a pivotal role in promoting angiogenesis in pancreatic cancer (15). Expression of phosphorylated STAT3 has been associated with increased VEGF-A expression in human pancreatic cancer tissues, and dominant-negative inhibition of STAT3 led to a significant reduction in tumor growth and vascularization in an experimental model (15). However, STAT3 is also required for the functionality of hypoxia-inducible factor-1α (HIF-1α), another transcription factor involved in the growth of pancreatic cancer (16, 17), underscoring a potential central role of STAT3 in pancreatic cancer.

With this background, we hypothesized that simultaneous targeting of these oncogenic signaling pathways (i.e., IGF-I/IGF-IR, IL-6/IL-6R, STAT3, and HIF-1α) could improve therapy of pancreatic cancer. Recently, the chaperone heat-shock protein 90 (Hsp90) has been recognized to be crucial for the stability and function of a wide variety of oncogenic kinases and signaling intermediates (18–20). Moreover, the transcription factors HIF-1α and STAT3 have been identified as being client proteins of Hsp90 (20, 21). Importantly, Hsp90 is not only expressed at a 2- to 10-fold higher level in cancer cells (22); recent studies also indicate that Hsp90 inhibitors elicit a significantly higher binding affinity to Hsp90 in cancer cells, compared with nonmalignant cells (23). As such, we proposed that the inhibition of Hsp90 may be valuable for targeting the above-mentioned signaling intermediates and transcription factors to reduce pancreatic cancer growth and angiogenesis. To date, the best characterized Hsp90 inhibitors are the geldanamycin derivates 17-allylamino-geldanamycin (17-AAG) and 17-(dimethylaminoethylamino)-17-demethoxygeldanamycin (17-DMAG), which are already being investigated in some phase I/II trials (24–26). Hence, we selected these inhibitors for our preclinical investigations. However, novel synthetic Hsp90 inhibitors are currently being developed, thus underlining the potential suitability of targeting Hsp90 for cancer therapy in the future (27, 28).

In the present study, we evaluated the effects of Hsp90 inhibition on proangiogenic signaling cascades in vitro and on tumor growth and metastasis in an experimental model with human pancreatic cancer cells in vivo. Importantly, we found that blocking Hsp90 directly inhibits IGF-IR phosphorylation and substantially impairs IL-6 signaling cascades, reduces HIF-1α and STAT3 activation, and diminishes VEGF-A expression in pancreatic cancer cells. Moreover, inhibition of Hsp90 strongly reduced pancreatic cancer growth and angiogenesis in an orthotopic tumor model.

Cell culture and reagents. The human pancreatic cancer cell line HPAF-II was obtained from American Type Culture Collection, and the metastatic L3.6pl cell line was kindly provided by Dr. I.J. Fidler (The University of Texas MD Anderson Cancer Center, Houston, TX). Cells were cultured in DMEM (Life Technologies) supplemented with 15% FCS and maintained in 5% CO₂ at 37°C as described (7, 29). Hsp90 inhibitors 17-AAG and 17-DMAG were purchased from Invivogen (Cayla-Invivogen). Radicicol (Sigma Aldrich), a non-geldanamycin Hsp90 inhibitor, was used for validating results on Hsp90 inhibition in vitro. 17-AAG was dissolved in DMSO, and respective concentrations of DMSO served as negative control. IGF-I and IL-6 were obtained from R&D Systems. Inhibitors to Erk1/2 (UO126) and phosphoinositide-3-kinase (PI3K)/Akt (LY290004) were obtained from Calbiochem (Merck).

3-(4,5-Dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide assay. To evaluate the cytotoxic effects of 17-AAG on tumor cells, L3.6pl and H-PAF-II cells were seeded into 96-well plates (1 × 10³ per well) and exposed to various concentrations of 17-AAG for 24 or 48 h. We used the 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) assay to assess cell viability as previously described (7).

Western blot analyses for activated signaling pathways. Experiments were done in triplicates at a cell density of 60% to 70% to reduce constitutive HIF-1α activation. Unless otherwise indicated, cells were incubated with 17-AAG (1 μmol/L) for 16 h before stimulation with either IGF-I (100 ng/mL) or IL-6 (50 ng/mL). Whole cell lysates were prepared as described elsewhere (7). Protein samples (50 μg) were subjected to Western blotting on a denaturing 10% SDS-PAGE. Membranes were sequentially probed with antibodies to indicated signaling intermediates. Antibodies to Hsp70, phospho-Akt^Ser473, phospho-Akt^Thr308, Akt, phospho-Erk^{Thr202/Tyr204}, Erk, phospho-STAT3^Tyr705, STAT3, phospho-STAT5^Tyr694, STAT5, phospho-FAK^Tyr925, FAK, phospho-Paxillin^Tyr118, Paxillin, phospho–IGF-1R^Tyr1131, and cAMP-responsive element binding protein (CREB) were purchased from Cell Signaling Technologies, and antibodies to β-actin and IGF-IRβ were from Santa Cruz Biotechnologies. Anti–phospho-IRS-1^Tyr612 was obtained from Calbiochem (Merck). Western blot analyses of tumor samples were done likewise after tissue lysis using the above extraction buffer as described elsewhere (30). For the analysis of IGF-IR expression in tumor specimens, immunoprecipitation of protein samples (800 μg) was done as described, using anti–IGF-IRβ antibody (Santa Cruz Biotechnologies; ref. 7).

Analyses for nuclear HIF-1α protein. Activation of HIF-1α was determined by Western blotting of nuclear protein extracts, which were prepared by using NucBuster Reagent Kit (Merck) according to the manufacturer's protocol. Nuclear protein extracts (75 μg) were subjected to SDS-PAGE (31). Membranes were probed with an anti–HIF-1α antibody (Novus Biologicals; Merck), and chemiluminescence detection was done thereafter (ECL, Amersham Bioscience).

Migration assays. To determine the impact of 17-AAG treatment on cancer cell motility in vitro, migration assays were done using modified Boyden chambers as described (32). Briefly, 5 × 10⁴ cells were resuspended in 1% FCS-DMEM and seeded into inserts with 8-μm filter pores (Becton Dickinson Bioscience). As chemoattractant, either 10% FCS or IGF-I (100 ng/mL) was used. After 48 h, cells were fixed, and migrated cells were stained (Diff-Quick reagent, Dade Behring). Cells that migrated through the filters were counted in four random fields, and average numbers were calculated.

ELISA for VEGF protein. To determine changes in VEGF, we used an ELISA kit specific to human VEGF-A (BioSource Europe) as previously described (31). Pancreatic cancer cells were plated at 40% to 50% density and incubated with or without 17-AAG and stimulated with either IGF-I or desferroxamine for 24 or 48 h. Analyses of culture supernatants were done according to the manufacturer's protocol.

Real-time PCR analyses. For real-time PCR (RT-PCR), total RNA was isolated using the TRIzol reagent (Invitrogen) and subsequently purified by ethanol precipitation. For each RNA sample, a 1-μg aliquot was reverse transcribed into cDNA using the Superscript II Kit (Qiagen). Primer pairs were as follows: VEGF165 (5′-GCACCCATGGCAGAAGGAGGAG; 3′-AGCCCCCGCATCGCATCAG), HIF-1α (5′-TACCATGCCCCAGATTCAGGAT; 3′-TCAGTGGTGGCAGTGGTAGTGG), IL-6 (5′-CCCAGTACCCCCAGGAGAAGA; 3′-GTTGGGTCAGGGGTGGTTATTG), IGF-IR β (5′-GTTGGGAAGGGGATCATTTT; 3′-CATGAAAACCATTGGCTGTG), and β-actin (5′-AGAGGGAAATCGTGCGTGAC; 3′-CAATAGTGATGACCTGGCCGT). Primers were optimized for MgCl₂ and annealing, and PCR products were confirmed by gel electrophoresis. RT-PCR was done using the LightCycler system and Roche Fast-Start Light Cycler-Master Hybridisation Probes master mix (Roche Diagnostics).

Transient RNAi for inhibition of STAT3. To specifically inhibit STAT3, siRNA_STAT3 (GCAACAGAUUGCCUGCAUUTT) was designed using an online RNAi-designer application.³

Pancreatic cancer cells were plated at 40% to 50% density and transiently transfected (25 mmol siRNA) using LipofectAMINE (Invitrogen). Knockdown effectiveness was tested by Western blotting for total STAT3.

Animal models. Eight-week-old male athymic nude mice (BALB/c ^nu/nu; Charles River) were used for experiments as approved by the Institutional Animal Care and Use Committee of the University of Regensburg and the regional authorities. In addition, experiments were conducted according to the Guidelines for the Welfare of Animals in Experimental Neoplasia published by The United Kingdom Coordinating Committee on Cancer Research. The effects of Hsp90 inhibition on the growth of human pancreatic cancer cells (HPAF-II) were investigated in a s.c. xenograft tumor model. Cancer cells (1 × 10⁶) were injected into the subcutis (right flank) of nude mice. Mice were randomized (n = 10 per group) and assigned to treatment groups. For in vivo experiments, the water-soluble Hsp90 inhibitor 17-DMAG was used because it is more potent and better suitable for i.p. injections than 17-AAG. I.p. injection of 17-DMAG (25 mg/kg, thrice a week) was started on day 5 when tumor became palpable. This dose was chosen based on the results from our previous studies, which showed that this dose is markedly below the reported MTD and well tolerated by mice (30). Tumor diameters were measured every other day, and tumor volumes were calculated (width² × length × 0.5). When the experiment was terminated, s.c. tumors were excised, weighed, and prepared for Western blot analyses.

The effects of 17-DMAG were evaluated in an orthotopic pancreatic cancer model using male athymic nude mice (BALB/c^nu/nu; Charles River), as described previously (7). In brief, 10⁶ human pancreatic cancer cells (L3.6pl) were injected into the pancreatic tail of mice. After implantation, tumors were allowed to grow 7 days before treatment was initiated. Mice were randomized into groups (n = 9-10 per group), receiving either vehicle (saline) or 17-DMAG (25 mg/kg, thrice a week) by i.p. injections. On day 28 after tumor cell inoculation, mice were sacrificed, and excised tumors were measured and weighed. For immunohistochemical analyses, tumors were either paraffin embedded or optimum cutting temperature (OCT) embedded.

Immunohistochemical analyses of tumor vascularization, cancer cell proliferation, and STAT3 phosphorylation. Multiple cryosections were obtained from tumors for all immunohistochemical analyses. CD31-positive vessel area was assessed using the rat anti-mouse CD31/platelet/endothelial cell adhesion molecule 1 (PECAM-1) antibody (PharMingen) and peroxidase-conjugated goat anti-rat immunoglobulin G (IgG; Jackson Research Laboratories) as previously described (7, 31). Antibody binding was visualized using stable diaminobenzidine. Images were obtained in four different quadrants of each tumor section (2 mm inside the tumor-normal tissue interface) at 40× magnification. The measurement of vessel area of CD31-stained vessels was done by converting images to grayscale and setting a consistent threshold for all slides using the ImageJ software (version 1.33; NIH). Vessel areas were expressed as pixels per high-power field (HPF; ref. 31). To determine the amount of proliferating tumor cells, mice received i.p. injections of bromodeoxyuridine (BrdUrd; Sigma Aldrich; 1 mg per mouse) 2 h before termination of animal studies. A commercially available BrdUrd detection kit (Becton Dickinson) was used to visualize BrdUrd uptake of cells in sections of tumors. Briefly, sections were incubated with anti-BrdUrd antibody solution, followed by streptavidin-conjugated horseradish peroxidase–linked goat anti-mouse IgG₂. Antibody binding was visualized by incubating slides in diaminobenzidine with the aid of hematoxylin counterstaining. BrdUrd-positive tumor cells were counted in four fields per tumor section at 20× magnification, and averages were calculated (31). Activation of STAT3 in tumor tissues was investigated by staining for phospho-STAT3 (1:100 dilution in 10% goat-serum/1% bovine serum albumin/PBS; Cell Signaling Technologies) on paraffin-embedded tissue sections according to our protocol described for mTOR staining (33).

Statistical analyses and densitometry. Statistical analyses were done using SigmaStat (Version 3.0). Results of in vivo experiments were analyzed for significant outliers using the Grubb's test for detecting outliers.⁴

Tumor-associated variables in in vivo experiments were tested for statistical significance using the Mann-Whitney U test for nonparametric data. The two-sided Student's t test was applied for analysis of in vitro data. All results are expressed as the mean ± SE unless stated otherwise in the legends. Densitometry was done on Western blots to quantify changes in expression or phosphorylation of signaling receptors and intermediates.

Inhibition of multiple signaling intermediates by blocking Hsp90 in pancreatic cancer cells. To identify the intracellular signaling pathways that are being affected by inhibition of Hsp90, we first investigated whether 17-AAG treatment leads to alterations in constitutively activated pathways that have been implicated in the regulation of angiogenic molecules such as VEGF. All signaling experiments were confirmed in a second cancer cell line (L3.6pl); however, data are representatively shown from experiments with HPAF-II cells throughout. Western blot analyses showed that 17-AAG strongly reduced constitutive phosphorylation of Akt, Erk1/2, and STAT3 in pancreatic cancer cells (Fig. 1A). In addition, the expression of total Akt seemed to be reduced by 17-AAG in cells after this time, a finding that has also been described for other cell lines (34). Because Akt and Erk1/2 pathways have previously been identified to be important for mediating IGF-I–induced up-regulation of VEGF in pancreatic cancer (7), we now hypothesized that the inhibition of Hsp90 would substantially interfere with IGF-I/IGF-IR signaling in this cancer entity. In preliminary experiments, IGF-I induced Akt and Erk1/2 in both pancreatic cancer cell lines (data not shown). Pretreatment of cells with 17-AAG clearly inhibited constitutive and IGF-I–mediated activation of Erk1/2 and Akt signaling substrates (Fig. 1B). Interestingly, phosphorylation of STAT3 was diminished independently of IGF-I stimulation, suggesting that STAT3 is not directly involved in an IGF-IR signaling cascade. Importantly, we also noted that IRS-1, which is upstream of Erk and Akt, was also responsive to 17-AAG in terms of impaired phosphorylation (Fig. 1B). As determined in subsequent experiments, inhibition of Akt and Erk1/2 by 17-AAG occurred in a dose-dependent manner, where 100 nmol/L 17-AAG notably diminished the phosphorylation of these substrates (Fig. 1C). We concluded from these experiments that blocking Hsp90 effectively impairs the downstream signaling of IGF-IR in pancreatic cancer.

We next sought to investigate whether the functionality of IGF-IR also depends on Hsp90 because geldanamycin and its derivates have been reported to elicit direct effects on phosphorylation and expression of various tyrosine kinases (20). Indeed, we detected a dose-dependent inhibition of IGF-IR phosphorylation after 16 h of 17-AAG exposure, but this effect was paralleled by a marked down-regulation of IGF-IR expression, which was determined by Western blotting for the IGF-IRβ because this component of IGF-IR contains the kinase domain (data not shown). Using a time course with 17-AAG, we subsequently determined that this down-regulation of IGF-IRβ occurs in a time-dependent fashion (Fig. 1D). Interestingly, we also noted that after 20 to 24 h, the precursor form of IGF-IRβ (Fig. 1D and E, arrow) had increased in parallel to IGF-IRβ down-regulation. To further investigate this finding, we used real-time PCR to determine changes in IGF-IRβ mRNA expression (precursor variant) upon Hsp90 inhibition. We found that 17-AAG treatment leads to a 2-fold increase in IGF-IRβ mRNA after 20 h, suggesting a potential compensatory mechanism of the cells. With regard to IGF-IR phosphorylation, at an early time point, 17-AAG treatment did not alter expression levels of IGF-IR, but markedly diminished IGF-I–mediated receptor phosphorylation (Fig. 1E). Similar to IGF-IR, IGF-I–mediated phosphorylation of IRS-1, a substrate of IGF-IR, was markedly diminished (Fig. 1E). Importantly, results on IGF-IR down-regulation and signaling interference were confirmed by using radicicol (5 μmol/L), a non-geldanamycin Hsp90 inhibitor (data not shown). Hence, results indicate that IGF-IR function is impaired earlier than its down-regulation occurs, suggesting that targeting Hsp90 elicits an immediate IGF-IR inhibitory effect.

Inhibition of Hsp90 disrupts IL-6/STAT3 signaling in pancreatic cancer. Interleukin-6 is a cytokine that has been implicated in the growth and angiogenesis of pancreatic cancer (8, 14). Because IL-6 signaling involves STAT3, a transcription factor that seems to be regulated by Hsp90 and that has been implicated in up-regulating VEGF in pancreatic cancer (15), we next determined IL-6–mediated activation of intracellular signaling pathways. Our results showed that IL-6 time-dependently induced phosphorylation of various substrates in pancreatic cancer cells (Fig. 2A). Focusing on the IL-6–mediated activation of STATs, treatment with 17-AAG effectively blunted the induction of STAT3 and STAT5 phosphorylation (Fig. 2B). Moreover, as IL-6 stimulation also led to an activation of Erk1/2 (which itself is regulated by Hsp90) in cancer cells, we subsequently ruled out the involvement of Erk1/2 in mediating the observed effects of 17-AAG on STAT3/STAT5 activation by using a specific signaling inhibitor to Erk1/2 (UO126; Fig. 2C). Similarly, selective inhibition of PI3K/Akt with the specific inhibitor LY290004 did not affect STATs phosphorylation upon IL-6 exposure (data not shown). Therefore, we conclude that inhibition of Hsp90 is capable of directly disrupting IL-6–mediated activation of STAT3 and STAT5 in human pancreatic cancer cells.

Effect of Hsp90 inhibition on IL-6 and hypoxia-mediated induction of HIF-1α. The transcription factor HIF-1α is known to be an important promoter of tumor growth and angiogenesis in various tumor entities, including pancreatic cancer (3, 7, 35, 36). Interestingly, HIF-1α is not only a client protein of Hsp90 (21), but is also required for forming a complex with STAT3 to facilitate the expression of target genes such as VEGF (16, 17). Hence, we hypothesized that Hsp90 inhibition will affect the functionality of HIF-1α through either direct inhibition or via blockade of STAT3 or both. Results showed that 17-AAG markedly diminished constitutive nuclear HIF-1α protein content in pancreatic cancer cells (Fig. 3A, top). In addition, induction of HIF-1α by chemical hypoxia (desferroxamine) was blunted by 17-AAG (Fig. 3A, middle). Similar to hypoxia, stimulation with IL-6 increased nuclear HIF-1α content. This response was reduced by inhibiting Hsp90 with 17-AAG (Fig. 3A, bottom). Moreover, inhibition of Hsp90 also directly diminished HIF-1α in terms of down-regulating its constitutive expression, which was determined by real-time PCR for HIF-1α mRNA (Fig. 3B).

To address the question whether 17-AAG–mediated inhibition of STAT3 is in part responsible for the observed reduction of HIF-1α upon Hsp90 inhibition, we used siRNA against STAT3 and investigated transiently transfected cells. STAT3_siRNA markedly down-regulated STAT3 expression in pancreatic cancer cells (Fig. 3C). Furthermore, STAT3_siRNA diminished both IL-6 and hypoxia-mediated induction of HIF-1α, suggesting that inhibition of STAT3, which occurs upon Hsp90 blockade, is a valid mechanism for reducing HIF-1α activity (Fig. 3C). In addition, we sought to investigate whether hypoxia itself could potentially induce IL-6 expression in pancreatic cancer cells, thereby contributing to an autocrine loop. Hypoxia led to an increase in IL-6 mRNA expression in pancreatic cancer, suggesting that HIF-1α and STAT3 may sustain an IL-6/HIF-1α/STAT3 autocrine activation loop (Fig. 3D). Importantly, stimulation of cells with recombinant IL-6 also led to an 8-fold increase in IL-6 mRNA after 24 h (data not shown). Together, these experiments provide evidence that an Hsp90 inhibitor can disrupt an HIF-1α/STAT3–mediated autocrine activation loop for IL-6 in pancreatic cancer through direct interference with both HIF-1α and STAT3 function.

Impact of Hsp90 inhibition on VEGF regulation. Because both IGF-IR and HIF-1α are important mediators for up-regulating VEGF-A in pancreatic cancer, we next investigated the effects of 17-AAG on VEGF-A expression and secretion. To address this issue, we first measured constitutive VEGF-A mRNA levels after 24 h of treatment with 17-AAG and found a 25% down-regulation of VEGF-A mRNA (data not shown). By ELISA analyses, IGF-I only slightly increased basal VEGF-A secretion in cells, which is likely due to the excessive amount of constitutive VEGF-A production, a finding that has been reported previously (7). However, 17-AAG significantly reduced VEGF-A secretion under both hypoxia (1% O₂) and IGF-I stimulation (Fig. 4A and B). We therefore conclude that inhibitors to Hsp90 could also be used to diminish VEGF-mediated angiogenesis in human pancreatic cancer.

Effect of Hsp90 inhibition on cancer cell migration. Our results up to this point indicate that expression HIF-1α and its function can be inhibited by Hsp90 blockade. In addition to regulating VEGF, HIF-1α also regulates cancer cell motility (37). Moreover, focal adhesion kinase (FAK), which is a primary mediator of cell motility, has been reported to be another client protein of Hsp90 (38). Hence, we determined the impact of 17-AAG treatment on FAK activation and observed an inhibition of constitutive FAK phosphorylation (Fig. 5A). Moreover, phosphorylated and total paxillin, which is a substrate of FAK, was substantially diminished (Fig. 5B). Consistent with this finding, constitutive and IGF-I–mediated tumor cell motility was significantly impaired by 17-AAG (Fig. 5C). Therefore, inhibition of Hsp90 could potentially reduce metastasis of pancreatic cancer through blocking migratory and invasive properties of cancer cells.

Effect of Hsp90 blockade on growth of pancreatic cancer. To estimate the effects of Hsp90 blockade on tumor growth and signaling pathway inhibition in vivo, we first used a s.c. xenograft model of pancreatic cancer (HPAF-II). Treatment with 17-DMAG significantly inhibited tumor growth (Fig. 6A). Importantly, as determined by Western blotting, IGF-IR was dramatically down-regulated in specimens derived from treatment groups (Fig. 6B). Moreover, phosphorylation of STAT3, the central mediator of an IL-6/STAT3/HIF-1α autocrine loop and essential cofactor for HIF-1α, was also markedly reduced in 17-DMAG–treated tumors, as well as phospho-Akt and phospho-Erk. Measuring Hsp70 expression was used to validate the efficacy of 17-DMAG in vivo, and therapy with 17-DMAG indeed increased Hsp70 protein in tumor samples. In contrast, substantial changes in expression or phosphorylation of FAK and paxillin were not detectable (data not shown). Next, growth-inhibitory and antiangiogenic properties of 17-DMAG were validated in an orthotopic pancreatic tumor model (L3.6pl; ref. 7). After 21 days, therapy with 17-DMAG led to a significant reduction in orthotopic tumor growth, as reflected by the final tumor weights (Fig. 7A). In addition, we detected a significant reduction in CD31-positive vessel area in tumors treated with 17-DMAG (Fig. 7B). Moreover, in accordance with measured tumor weights, numbers of proliferating tumor cells were significantly diminished in the 17-DMAG therapy arm (Fig. 7C). Similar to s.c. tumors, activation of STAT3 was substantially diminished in tumors treated with 17-DMAG (Fig. 7D). Together, these results indicate that IGF-IR and STAT3 are important targets of Hsp90 inhibitors in pancreatic cancer, which could be used to effectively inhibit growth and vascularization of pancreatic cancer.

In this study, we show that blocking Hsp90 directly inhibits IGF-IR function and IL-6–mediated proangiogenic signaling cascades in pancreatic cancer cells and also harbors the potential to reduce cancer cell invasiveness. We additionally identified a novel IL-6/STAT3/HIF-1α autocrine activation loop in pancreatic cancer cells that was disruptable by blocking Hsp90. The therapeutic suitability of targeting Hsp90 was subsequently validated in an orthotopic tumor model, where the Hsp90 inhibitor 17-DMAG strongly reduced pancreatic cancer growth and tumor vascularization. Our results therefore provide a novel multifactorial rationale for using Hsp90 inhibitors in therapy concepts for treating pancreatic cancer.

The importance of the IGF-IR system in pancreatic cancer growth and angiogenesis is apparent from previous studies (6, 7, 9); thus, IGF-IR has become an interesting target for cancer therapy. However, the development of specific and effective antagonists to this receptor system has been challenging and is currently ongoing. Importantly, we now have identified that Hsp90 inhibitors do elicit a direct effect on IGF-IR function in terms of reduced receptor phosphorylation upon stimulation with its ligand. This direct inhibitory effect precedes a consecutive down-regulation of IGF-IR itself, which occurs in a time- and dose-dependent manner. Hence, we propose that targeting Hsp90 elicits acute (inhibition of phosphorylation) and chronic (down-regulation) inhibitory effects on the IGF-IR. Interestingly, we also noted that Hsp90 inhibition leads to an initial drop in IGF-IRβ mRNA expression, followed by a moderate increase in IGF-IRβ mRNA, suggesting a potential compensatory response. This finding could explain observed increases in the IGF-IRβ precursor form in Western blot analyses. In addition, we speculate that this increase of the precursor form could also be a result of an Hsp90-inhibitor–mediated impairment of the pro-convertase furin/PC5 (39). To date, only the effect of a IGF-IR down-regulation by Hsp90 inhibitors has been reported in a few studies, whereas the impact on IGF-IR phosphorylation was not addressed in any of these studies. Bagatell et al. (40) observed a down-regulation of IGF-IR in various osteosarcoma cell lines upon treatment with geldanamycin, one of the first described Hsp90 inhibitors that is less potent compared with novel derivates. In a second study, Terry et al. showed that the geldanamycin-derivate 17-AAG diminished the expression of IGF-IR in one of two synovial carcinoma cell lines (41), and the authors concluded that particularly, IGF-IR–overexpressing cancer cells do react to a Hsp90-inhibiting therapy. Together, these findings make inhibitors to Hsp90 very attractive to be used in therapy concepts for treating pancreatic cancer, as IGF-IR is frequently overexpressed and constitutively active in this cancer (6, 7, 9).

Apart from the IGF-IR system, we examined the possibility that inhibition of Hsp90 could affect pancreatic cancer growth and angiogenesis through yet another mechanism. The mechanism we studied involved IL-6 and its potential to promote angiogenesis through up-regulating VEGF-A expression (14, 42, 43). Previously, Tang et al. (8) have described an IL-6–mediated up-regulation of VEGF-A in various pancreatic cancer cell lines in vitro, and Masui et al. (14) have correlated IL-6 receptor expression with VEGF expression in human pancreatic cancer specimens. Furthermore, it has been shown that IL-6 signaling in cancer cells may involve the activation of the transcription factor STAT3 (44). Important for our study, STAT3 is essential for the functionality of the HIF-1α complex, and both transcription factors seem to be Hsp90 client proteins (16, 17). Consistent with these associations, we showed in the present study that IL-6–mediated activation of STAT3/5 could be effectively blocked in pancreatic cancer cells by 17-AAG. Moreover, stimulation with IL-6 led to an increase in nuclear HIF-1α content and likewise was blunted by 17-AAG treatment. Interestingly, we also identified a novel IL-6 autocrine activation loop because hypoxia and recombinant IL-6 itself led to an increase in IL-6 mRNA levels in pancreatic cancer cells, which has not been described previously. Similar to the IGF-IR system (6, 7), this autocrine IL-6/STAT3/HIF-1α loop may contribute to the aggressive course of this disease through sustaining a constitutive STAT3/HIF-1α activation and up-regulating VEGF. We propose, therefore, that targeting Hsp90 may be effective for disrupting both the IGF-IR system and an IL-6/STAT3/HIF-1α autocrine activation loop in pancreatic cancer. In addition, we found that 17-AAG also inhibited the phosphorylation of STAT5. Activation of this transcription factor has been implicated in promoting the transformed phenotype in cancers, as well as in mediating tolerance to hypoxia (45, 46), thus representing another interesting target for cancer therapy.

The potential growth-inhibitory and antiangiogenic effects of Hsp90-targeted therapy were investigated in a s.c. xenograft tumor model and in an orthotopic tumor model with pancreatic cancer cells. Our study is the first to show that Hsp90 inhibitors, such as 17-DMAG, potently reduce tumor growth and vascularization of pancreatic tumors. Importantly, two major regulators of pancreatic cancer growth and angiogenesis, i.e., IGF-IR and STAT3, can be effectively inhibited by such Hsp90 inhibitors in vivo. This finding may also be of particular importance for clinical trials, as efficacy of Hsp90-targeted therapy could potentially be monitored by measuring these client proteins in fine-needle biopsies from pancreatic tumors, as suggested by a recent study (47). In our view, the measuring activity of such client proteins is superior to determining the increase of Hsp70 as a biomarker because reactive Hsp70 increases do not allow direct conclusions on Hsp90 activity/function. Moreover, in contrast to reported studies that investigated antineoplastic efficacy of Hsp90 inhibitors on malignancies other than pancreatic cancer, we used only half of the maximum tolerated dose (48, 49). This aspect will be crucial for minimizing potential side effects of Hsp90-targeted therapy, and we now provide evidence that the concept of targeting multiple oncogenic pathways is efficient and does not require high-dose Hsp90 inhibitors.

To date, the role of Hsp90 inhibitors in metastasis has not been clear, since Price et al. (50) recently showed that geldanamycin and its derivates could promote bone metastases formation in a model of breast cancer. In pancreatic cancer, we now provide evidence that the inhibition of Hsp90 inhibits promigratory molecules (FAK, paxillin) and impairs migration of pancreatic cancer cells. However, we could not prove that FAK is indeed inhibited in tumor tissues by 17-DMAG therapy (administered thrice a week), which is probably due to a fast recovery of these proteins, as suggested by other studies (40). A regimen with daily 17-DMAG therapy could have been even more efficacious. Nevertheless, mice in the 17-DMAG treatment group elicited reduced lymph node metastasis (data not shown), suggesting that metastatic properties of pancreatic cancer cells can be functionally impaired in vivo.

In conclusion, our study shows that the inhibition of Hsp90 impairs IGF-IR function and disrupts an IL-6/HIF-1α/STAT3 autocrine activation loop in pancreatic cancer. These effects are associated with a significant reduction in tumor growth and vascularization in an orthotopic pancreatic cancer model. Hsp90 inhibitors may therefore be a valuable addition to molecular targeted therapy concepts for the treatment of pancreatic cancer.

The authors thank Christine Wagner, Corina de Sousa, and Kathrin Stengel for excellent technical assistance.