Background
Hypoxia inducible factor (HIF) is a master regulator of the hypoxic response and plays a critical role in the development and progression of numerous solid cancers [
1,
2]. HIF functions as a heterodimeric transcription factor composed of an oxygen regulated α-subunit and a constitutively expressed β-subunit (or ARNT). HIF activity is tightly regulated by oxygen tension wherein its activity is restrained under oxygenated conditions via von-Hippel Lindau (VHL) ubiquitin ligase mediated degradation of the α subunit [
3]. In contrast, tumor hypoxia facilitates HIF-α stabilization, dimerization, and transcriptional activation. HIF regulates a multitude of genes that contribute to pro-tumorigenic processes including invasion, angiogenesis and therapeutic resistance [
2,
4‐
6]. Importantly, inhibition of HIF function suppresses tumor formation and progression, and restores treatment sensitivity, highlighting HIF as a clinically relevant therapeutic target [
1,
7].
Clear cell renal cell carcinoma (CCRCC) tumors are highly vascularized and among the most lethal kidney tumors [
8]. CCRCC, with its defined loss of VHL function and resulting constitutive HIF expression and activity, is a useful model to decipher the role of HIF in cancer progression and to evaluate HIF targeting strategies. Although the sufficiency of HIF for CCRCC remains somewhat controversial [
9], HIF is a major participant in CCRCC within the context of VHL loss [
10‐
13]. Of the two main pro-tumorigenic HIF-α isoforms, HIF-2α elicits tumor formation in CCRCC xenograft models [
10,
14] and appears to be more commonly upregulated in CCRCC relative to HIF-1α [
4]. However, HIF-1α driven CCRCC xenograft models have also been documented [
15], as well as compensatory mechanisms between the two isoforms [
16]. Therefore, the targeting of both HIF isoforms may represent the most effective therapeutic approach. In spite of this, few studies have addressed the ability of candidate agents to target both isoforms.
A number of generalized HIF targeted approaches have been employed, including modulation of HIF expression, transcription, translation, dimerization, transactivation, and stability [
17‐
23]. Small molecule inhibitors of the chaperone heat shock protein 90 (Hsp90) represent a growing class of clinically utilized anti-tumorigenic agents that have been collectively exploited as an alternative means of targeting HIF-α, given their shared ability to disrupt the ATP dependent chaperone activity of Hsp90 and block the protein folding of respective Hsp90 clients. HIF is an Hsp90 client protein [
24] and we, and others, have shown that perturbing Hsp90 function with geldanamycin (GA) and small molecule derivatives promotes HIF-1α and HIF-2α protein degradation and suppression of transcriptional activity [
25‐
27]. Importantly, Hsp90 targeted approaches bypass the requirement for both VHL and oxygen, instead utilizing the ubiquitin ligase RACK1 [
25,
28]. Therefore, these agents hold promise in tumor environments where VHL function is compromised, as in CCRCC or tumor hypoxia. In support of this premise, the Hsp90 inhibitors GA, 17-(allylamino)-17-demethoxygeldanamycin (17-AAG or Tanespimycin) and 17-dimethylaminoethylamino-17-demethoxygeldanamycin (17-DMAG or Alvespimycin) demonstrate anti-tumorigenic and anti-angiogenic properties in both
in vitro and
in vivo animal models, due in part to their ability to inhibit HIF function [
29‐
33]. However, despite the promising pre-clinical actions of these inhibitors, clinical trials with 17-DMAG have been relatively unsuccessful for CCRCC and other solid tumors [
34‐
36]. These failures highlight the critical need to further evaluate the effects of Hsp90 targeting agents upon HIF dependent signaling and angiogenesis in CCRCC and other cancers.
Since the advent of 17-AAG, numerous Hsp90 inhibitors exhibiting enhanced potency and diminished toxicity have been developed [
37‐
41], leaving open the possibility that these next generation agents may demonstrate increased potency and efficacy
in vivo. In addition to these direct Hsp90 inhibitors, histone deacetylase (HDAC) inhibitors, such as the clinically utilized LBH589 (Panobinostat), indirectly inhibit Hsp90 through protein hyperacetylation, leading to loss of chaperone function [
42‐
44] and comparable inhibition of a subset of Hsp90 client proteins [
45‐
47]. Similar to Hsp90 inhibitors, HDAC inhibitors ablate HIF levels and activity in several models [
42,
45,
48‐
50]. Although one report examined the effects of HDAC inhibition upon HIF-1 expression in CCRCC cells [
50] and another utilized LBH589 in combination with rapamycin [
15], an analysis of these agents upon HIF-2 expression or activity is lacking, an important oversight given that a majority of CCRCC tumors preferentially express HIF-2 [
4]. Moreover, no reports have directly compared the effects of ATP based competitive Hsp90 inhibitors with HDAC inhibitors such as LBH589 with respect to their respective ability to suppress the activity of HIF isoforms.
In this study, we utilized a prototypic next generation Hsp90 inhibitor, EC154, representing a novel synthetic compound with enhanced potency and diminished toxicity relative to 17-AAG. EC154 is a compound with improved properties (potency, pharmacokinetic and pharmacodynamic) over BIIB021, the first oral synthetic Hsp90 inhibitor to reach clinical trials [
51]. EC154 binds to the Hsp90 ATP binding site in a manner similar to EC144 [
52]. Towards the goal of identifying more potent HIF targeting agents, we evaluated the ability of 17-AAG, EC154, and LBH589 to suppress HIF activity and angiogenic potential in CCRCC cells. We show herein that although these agents exert inhibitory effects against HIF activity, they exhibited differential time and dose dependent responses not consistently linked to relative HIF protein expression. Our results therefore provide valuable guidance on the use of these agents as HIF inhibitors and highlight the biological complexity of Hsp90 inhibition upon suppression of HIF activity and downstream targets. Further, our results suggest that the evaluation of HIF expression in tissues as a surrogate for therapeutic efficacy may not suffice as a reliable indicator of HIF suppression.
Methods
Reagents and antibodies
The Hsp90 inhibitors 17-(Allylamino)-17-demethoxygeldanamycin (17-AAG) and EC154 were provided by the NCI Developmental Therapeutics Program and Biogen IDEC, respectively. The HDAC inhibitor LBH589 was provided by Dr. Peter Atajda (Novartis). HIF-1α (100-105) and HIF-2α (100-122) antibodies were from Novus Biologicals. Additional antibodies included GAPDH (9295) from Sigma; RACKI (17754) and topoisomerase II (13059) from Santa Cruz; P-ERKT202/Y204 (4370), ERK (4695), P-srcY416 (2101), src (2108) from Cell Signaling; and P-FAKY397 (44-624G) and FAK (AHO-0502) from Invitrogen.
Cell culture
UMRC2 and 786-O cells were provided by Dr. M.I. Lerman and Dr. R. Klausner (NCI, National Institutes of Health). UMRC2-VHL replaced cells were made as previously described [
25] and 786-O VHL-replaced cells were provided by Dr. W.G. Kaelin (Dana-Farber Cancer Institute). All CCRCC cell lines were maintained in buffered Dulbecco's modified Eagle's medium (Thermo Scientific) supplemented with 10% fetal bovine serum (Gibco), L-glutamine (Thermo Scientific), and penicillin/streptomycin (Mediatech). Human umbilical vascular endothelial cells (HUVEC) were purchased from Invitrogen and maintained in Medium 200/LSGS (Gibco). All cells were maintained at 37°C and 5% CO
2
Western blots
Nuclear preparations were prepared as previously described [
25]. Briefly, cells were washed in phosphate-buffered saline, collected using a hypotonic buffer (20 mM HEPES, 5 mM MgCl
2, 5 mM NaCl, 1 mM EDTA, pH 7.9) containing protease inhibitors (Roche), nuclear pellets were collected following addition 10% Nonidet P-4, lysed in hypertonic buffer (20 mM HEPES, 400 mM NaCl, 1 mM EDTA) containing protease inhibitors, and supernatant collected. For whole cell lysates, cells were washed with phosphate-buffered saline, incubated in lysis buffer (20 mM Tris pH 7.5, 150 mM NaCl, 1%NP40, 1 mM EDTA) containing protease inhibitors, and supernatant collected. All protein concentrations were determined by BCA method (Thermo Scientific).
qRT-PCR analysis
CCRCC cells were treated with 17-AAG, EC154, LBH589 or DMSO solvent control. Total RNA was extracted with the RNeasy Kit (Qiagen, Valencia, CA). Total RNA (1 μg) was used for reverse transcription with SuperScript® III First-Strand Synthesis SuperMix for qRT-PCR (Invitrogen,) and resulting cDNA was used as a template for quantitative Real-time PCR analysis. The ratio of GLUT1, VEGF, LOX1 and OCT-4 to α-tubulin was measured with real-time-quantitative PCR (iQ5 Multicolor Real-Time PCR Detection System, Bio-Rad Laboratories, Hercules, CA) using iQ5 optical system software. The reaction mixture contained 12.5 μl SYBR green mix (Bio-Rad), 2 μl of mixed primers, and 2 μl cDNA template. The final volume was adjusted with H2O to 25 μl. Annealing temperature was 55°C for all reactions. Fluorescent products were measured by a single acquisition mode after each cycle. Primers are as follows: α-tubulin: F-AACGTCAAGACGGCCGTGT, R-GACAGAGGCAAACTGAGCAC; GLUT1: F-GTGGGCATGTGCTTCCAGTA, R-ACAGAACCAGGAGCACAGTGAA; VEGF: F-AGGCCAGCACATAGGAGAGA, R-TTTCCCTTTCCTCGAACTGA; CAIX: F-GGGTGTCATCTGGACTGTGTT, R-CTTCTGTGCTGCCTTCTCATC; qLOXI: F-ATGAGTTTAGCCACTTGTACCTGCTT, R-AAACTTGCTTTGTGGCCTTCA; OCT-4: F-GACAACAATGAGAACCTTCAG GAGA, R-CTGGCGCCGGTTACAGAACCA.
Lentiviral infection and luciferase reporter assays
The Cignal Lenti-HIF Reporter system (SABiosciences) was used to stably transduce CCRCC cells with constitutively expressed renilla luciferase and HIF-regulated firefly luciferase constructs. UMRC2 and 786-0 cells were grown to 60-70% confluence and infected for 24 h according to the manufacturer's specifications. Cells were placed under puromycin selection and positive clones were expanded. For analysis of HIF activity, cells at 70% confluency were treated for 16 h with or without inhibitors, lysed, and dual luciferase activity analyzed (Promega). Experiments were performed in triplicate and arbitrary values normalized to renilla luciferase levels.
Immunoassays
UMRC2 and 786-O cells were pre-treated for 4 h with the indicated compounds in low serum (3% FBS) medium. Cells were rinsed with PBS and then incubated with freshly prepared treatments in similar medium for 16 h under or normoxia or hypoxia (1% O2). Conditioned medium was collected following brief centrifugation and whole cell lysate was collected as described. VEGF and uPA concentrations were measured by ELISA (R&D Systems) according to the manufacturer's instructions, and normalized to total protein concentration of the conditioned medium.
HUVEC cells were serum starved overnight (0.1% M200) and plated (15,000 cells/well) in 96-well plates pre-coated with 50 μL of phenol-red free growth factor reduced Matrigel (BD Biosciences) with the indicated treatments. CCRCC conditioned medium was collected and added to HUVEC cells for 6 h. Effects on angiogenesis were determined by quantifying branch points in 6 replicate wells (1 field per well; 40X) for each treatment and are presented as percent of untreated control with standard deviation.
Cell migration
Cell migration was assessed using Boyden chambers (0.8 μm; BD Biosciences). Cells were plated into the upper chamber in reduced serum DMEM (0.1%) with complete DMEM in the lower chamber and indicated agents added to both chambers. At 24 h, inserts were washed with PBS, the non-motile cells removed with cotton swabs, and the migratory cells fixed in formaldehyde, visualized with 0.1% crystal violet, and counted. The data presented represent the mean value from 4 replicates per treatment.
Cell viability
CCRCC cells (5,000) were plated in 96-well plates, treated with 17-AAG, EC154, LBH589 or vehicle for 16 h and assessed for cell viability using the Cell-Titer Blue Reagent as per the manufacturer's instructions (Promega).
Electrical Cell Substrate Impedance Sensing Assay (ECIS)
Changes in endothelial cell monolayer permeability were determined using the well established method of measuring electrical impendence [
53]. HUVECs were seeded onto ECIS 8W10E+electrode arrays (Applied Biophysics, Troy, NY) precoated with human plasma fibronectin (Invitrogen) at 100 μg/mL in 0.15 M NaCl, 0.01 M Tris, pH 8.0. Transendothelial electrical resistance (TEER), an index of endothelial cell barrier function, was measured using an ECIS Model 1600 instrument (Applied Biophysics, Troy, NY). Cells were allowed to form a confluent monolayer and cell barrier until a plateau was reached (~3 days). In order to evaluate the ability of 17-AAG to reverse the effects of agents known to disrupt barrier function, culture medium was replaced with fresh EGM-2 containing the barrier disruptor, VEGF (50 ng/mL) in the absence or presence of 17-AAG (1 μM), and impedance was measured every 5 min at 15 kHz frequency. In order to evaluate the effects of CCRCC conditioned medium (CM) on endothelial cell barrier function, culture medium was replaced with CM collected for 24 h from 1 × 10
6 UMRC2 or 786-0 cells in the absence or presence of 17-AAG (1 μM), EC154 (100 nM), or LBH589 (100 nM). HUVEC-CM was used as a control for possible nutrient depletion. The traces shown represent the mean of duplicate wells for each treatment.
Statistical analysis
Statistical significance was determined using one-way ANOVA followed by one-tailed student's T-tests. IC50 values were calculated using SigmaPlot10 (Systat) wherein curves were fitted to either a 3-parameter logistical or 4-parameter sigmoidal plot (R2 = 0.695-0.996).
Discussion
Given the well established role of HIF in CCRCC, and its seemingly universal involvement in the progression of diverse solid tumors, it is critically important to evaluate the HIF targeting ability of next-generation Hsp90 inhibitors. However, surprisingly few studies have examined this particular aspect of these agents. The results from this study indicate that standard endpoints for HIF function may be less informative than previously indicated, and that the apparent complexity of inhibitor effects requires a more functionally based and integrative approach. One example of this complexity is demonstrated by our findings that although all three compounds significantly reduced HIF-1α protein levels, consistent with previous reports [
25‐
27,
45], 17-AAG and EC154 did not appreciably diminish HIF-2α levels, and surprisingly, LBH589 increased HIF-2α expression. To our knowledge, this is the first report to examine the effect of an HDAC inhibitor on HIF-2α levels, and more significantly, to demonstrate an HDAC dependent increase in protein expression. It remains unclear whether the mechanism of LBH589 mediated increase of HIF-2α protein may be due to changes in protein degradation or synthesis. Further, the requirement of RACK1 for Hsp90-inhibition mediated HIF-2α degradation has not been established. The differential response of the HIF isoforms is not without precedent, in that isoform specificity for HIF-1α and HIF-2α has been demonstrated during oxygen and VHL independent destruction pathways [
86,
87]. However, of greater relevance to this study is that the LBH-589 mediated increase in HIF-2α expression did not elicit a comparable increase in HIF-2α activity.
Our findings reinforce the theme that HIF expression and activity may be uncoupled events, a notion previously observed following treatment with proteasomal inhibitors [
59]. First, although LBH589 increased HIF-2 expression, all other parameters of HIF activity support an overall suppressive effect. Given that HIF activity is regulated by acetylation, it is very possible that the increase in HIF acetylation by LBH-589 is inhibitory. Further, increased acetylation has also been shown to inhibit Hsp90 chaperone activity [
44], which may further suppress HIF-2 activity. Second, all three drugs diminished HIF-1α and HIF-2α dependent transcription within 2 h of treatment, prior to decreases in protein levels [
25,
88], and data not shown. Although EC154 and 17-AAG demonstrated approximately equivalent kinetics of suppression, exposure of cells to low concentrations of EC154 (10-fold lower levels than the experimental 100 nM) did not elicit detectable increases in src phosphorylation, indicating an improved profile compared with 17-AAG.
Our data highlight the complexity of drug targeting, in that the drug mediated suppression of HIF target genes varied dynamically with time and dose, and differential sensitivity was observed in a transcript and cell context dependent manner. This complexity is not surprising given that additional modifiers may contribute to the net effects of Hsp90 inhibition. For example, NF-kB promotes HIF-1α transactivation in an acetylation sensitive manner [
89] and coordinates suppressive effects upon NF-kB, and Hsp90 may contribute to the robust inhibition of HIF activity observed with LBH589. Moreover, HDAC inhibition at early time points has also been demonstrated to disrupt the p300:HIF complex, a critical component of HIF transactivation, resulting in decreased HIF transcriptional activity [
88]. Adding to this complexity, HIF directly interacts with, and is regulated by, HDACs [
90] illustrating several nodal points by which Hsp90 and HDAC inhibitors may distinctly impact upon HIF function.
To more comprehensively evaluate the relative efficacy of these agents against specific metrics, we examined key anticancer properties beyond HIF transcription. Sub-nanomolar concentrations of 17-AAG and EC154 profoundly suppressed CCRCC cell motility, whereas higher concentrations were required to diminish CCRCC-mediated HUVEC tubule formation. Although LBH589 demonstrated the most potent suppression of both
VEGF transcription and secreted protein, it only modestly prevented tubule formation in 786-O. Our tubule formation results support the notion that Hsp90 inhibiting agents, in part via their suppressive effects upon HIF, may be utilized clinically to reduce the vascularity associated with CCRCC and other solid tumors [
31,
67]. Although LBH589 did not restore barrier function, another report demonstrates that this agent, in combination with rapamycin in a preclinical model, was effective in reducing CCRCC angiogenesis [
15]. A recent report found that pre-incubation with the HDAC6 inhibitor tubacin prevents thrombin induced barrier damage [
91], suggesting that the effects of HDAC inhibition upon vessel permeability could vary greatly depending upon a variety of factors. Therefore, the ECIS assay may be a useful approach to more fully interrogate the ability of LBH589, and other HDACIs, to restore endothelial cell function alone and in combination with currently utilized anticancer agents. The ability of these agents to preserve endothelial cell integrity could have significant effects on clinical outcome and warrants further attention. Clinically, vascular permeability can promote tumor cell TEM and metastasis [
82] and poses a further challenge in diminishing the delivery and efficacy of chemo- and radio-therapies [
92]. Our results suggest that these agents may have utility in 'normalizing' the tumor vasculature, an effect that improves chemo- and radio-therapy [
92]. Although not formally demonstrated for Hsp90 inhibitors, 17-AAG and radicicol have been shown to improve endothelial barrier formation and rescue barrier in the presence of damaging agents including thrombin, VEGF, and phorbal ester [
84,
85]. In addition, we previously demonstrated that both GA and 17-AAG potentiate the radiation response of cervical tumor cells
in vitro and
in vivo [
29]. Our data suggest that EC154 exhibits a similar capacity as 17-AAG in terms of its ability to restore barrier function, and may therefore represent a viable clinical candidate.
Conclusions
In sum, our transcriptional results indicate that Hsp90 inhibition may affect HIF-dependent gene transcription through at least two mechanisms; by direct inhibition of transcriptional activity at early time points, and by promotion of HIF-1α degradation at later time points. This further highlights the complexity of Hsp90 inhibitor effects and weakens the notion that variations of HIF-α isoform stability is a primary reason for the clinical failure of these agents [
34]. Importantly, our results also highlight that HIF expression is not necessarily a reliable surrogate for interrogating the HIF targeting efficacy of Hsp90 inhibitors. This is a particularly salient finding in light of recent efforts to improve
in vivo and intratumoral imaging of Hsp90 client proteins as a surrogate for monitoring inhibitor activity [
93‐
97]. Moreover, our results further indicate that levels of HIF regulated cytokines such as VEGF may also be unreliable surrogates for HIF activity, demonstrated by the observed differential effects of Hsp90 inhibition upon intracellular and secreted VEGF. This finding further highlights the challenge in choosing appropriate biomarkers, in that the histochemical analysis of tumor tissues for VEGF expression would indicate less robust drug dependent effects compared with an analysis of secreted levels of the same protein. Our findings highlight the need to integrate multiple, diverse, and functional endpoints to more appropriately discern the effects of Hsp90 inhibition upon HIF function. With more than a dozen new Hsp90 inhibitors currently under clinical evaluation [
98], there is a critical need to establish relevant and reliable surrogates and biomarkers for evidence of Hsp90 inhibition. The identification of these readouts and their subsequent incorporation into clinical trials will provide useful indicators of positive response. Our findings offer useful metrics and considerations to guide the preclinical and clinical evaluation of conventional and novel Hsp90 inhibitors.
Competing interests
This work was funded in part by a commercial grant from Biogen Idec.
Authors' contributions
JEB participated in the design of the study, performed immunoblots, immunoassays, cell migration, tubule formation, cell viability and ECIS experiments, statistical analysis, and drafted the manuscript. SP carried out the qRT-PCR studies, lentiviral infections, and reporter assays. UG carried out studies pertinent to the activity status of pro-motility effector proteins. MWH participated in the cell migration studies. SW and KA participated in the design, execution, and analysis of the ECIS experiments. KL participated in the design and coordination for the study and provided EC154. JSI conceived of the study, participated in its design and coordination, and edited the manuscript. All authors read and approved the final manuscript.