Arylsulfonamide KCN1 inhibits in vivo glioma growth and interferes with HIF signaling by disrupting HIF-1α interaction with co-factors p300/CBP

Abstract

Purpose

The hypoxia inducible factor-1 (HIF-1) plays a critical role in tumor adaptation to hypoxia, and its elevated expression correlates with poor prognosis and treatment failure in cancer patients. In this study, we determined whether 3,4-dimethoxy-N-[(2,2-dimethyl-2H-chromen-6-yl)methyl]-N-phenylbenzenesulfonamide, KCN1, the lead inhibitor in a novel class of arylsulfonamide inhibitors of the HIF-1 pathway, had anti-tumorigenic properties in vivo and further defined its mechanism of action.

Experimental Design

We studied the inhibitory effect of systemic KCN1 delivery on the growth of human brain tumors in mice. To define mechanisms of KCN1 anti-HIF activities, we examined its influence on the assembly of a functional HIF1α/HIF1β/p300 transcription complex.

Results

KCN1 specifically inhibited HIF reporter gene activity in several glioma cell lines at the nanomolar level. KCN1 also downregulated transcription of endogenous HIF-1 target genes, such as VEGF, Glut-1 and carbonic anhydrase 9, in an HRE-dependent manner. KCN1 potently inhibited the growth of subcutaneous malignant glioma tumor xenografts with minimal adverse effects on the host. It also induced a temporary survival benefit in an intracranial model of glioma but had no effect in a model of melanoma metastasis to the brain. Mechanistically, KCN1 did not down-regulate levels of HIF-1α or other components of the HIF transcriptional complex; rather, it antagonized hypoxia-inducible transcription by disrupting the interaction of HIF-1α with transcriptional co-activators p300/CBP.

Conclusions

Our results suggest that the new HIF pathway inhibitor KCN1 has antitumor activity in mouse models, supporting its further translation for the treatment of human tumors displaying hypoxia or HIF overexpression.

Introduction

Hypoxia is a microenvironmental condition that is prevalent in solid tumor development, largely due to inadequate vascularization and rapid proliferation of tumor cells (1-3). To counter the detrimental effects of hypoxia, tumor cells activate a range of adaptive molecular mechanisms that play a critical role in all hallmarks of cancer (4). These include switching from oxidative phosphorylation to anaerobic glycolysis, angiogenesis, increased cell migration potential, and genetic alterations that prevent hypoxia-induced apoptosis. A family of heterodimeric transcription factors termed Hypoxia Inducible Factors (HIFs) governs the primary transcriptional response to hypoxia. HIFs consist of one of HIF-1α, 2α, or 3α (the O₂-regulated subunits) and the constitutively expressed HIF-1β (5). Under normoxic conditions, α subunits are hydroxylated by a family of prolylhydroxylases, ubiquitylated in a Von Hippel-Lindau protein-dependent manner, and degraded in the proteasome (6). Under hypoxic conditions, α subunits are stabilized, translocate into the nucleus where they interact with the HIF-1β subunit, recruit co-activators p300/CBP, and regulate (HIF-1 and 2 positively, HIF-3 negatively) over 100 target genes via binding to specific DNAs sequences termed hypoxia-responsive elements (HRE) (7).

CBP and p300 are homologous transcriptional co-activators, which act as a bridge linking DNA-binding transcription factors to the basal transcriptional machinery (8, 9). p300/CBP possess strong histone acetyltransferase activity that regulates remodeling of local chromatin structures and makes DNA more accessible to other regulators (8). The interaction between HIF-1α and p300/CBP, mediated by the C-terminal activation domain (CAD) of the former and the cysteine-histidine rich 1 (CH1) domain of the latter (10), is physiologically regulated via O₂-dependent hydroxylation of N803 in CAD by Factor Inhibiting HIF-1 (FIH-1) (6). Recently, a weaker interaction between the HIF-1α N-terminal activation domain and p300/CBP CH3 was also reported (11). The critical role of p300/CBP in HIF function has been established by showing that blockade of the HIF-1α - p300/CBP interaction markedly attenuated HIF activity (12)

The close relation of HIF-activated gene products with tumor progression/metastasis identifies HIF as an attractive therapeutic target. Several studies have already established that inhibition of the HIF pathway can inhibit malignant characteristics in a number of cancers (13, 14) and several small molecule inhibitors of HIF signaling have already been described (15-19). In addition, many anti-cancer compounds used in the clinic or in preclinical development were found to inhibit the HIF pathway indirectly (20-24). Despite this, new inhibitors of the HIF pathway, preferentially with defined and/or novel mechanism of action, need to be identified, and it is currently too early to determine which agent will have the best anti-tumor efficacy and safety profile.

To identify novel chemotypes with anti-HIF pathway activity, we previously performed a cell-based screen to identify small molecule inhibitors of HIF transcriptional activity in a combinatorial library (>10,000 compounds) built upon a 2,2-dimethyl-2H-chromene scaffold (25). In this library, we have identified arylsulfonamides as a novel chemotype with high nano-to-low micromolar (IC₅₀) HIF inhibitory activity (26). Here, we demonstrate that the lead compound identified in the screen, 3,4-dimethoxy-N-[(2,2-dimethyl-2H-chromen-6-yl)methyl]-N-phenylbenzenesulfonamide (KCN1) inhibits HIF transcriptional activity through the disruption of the interaction between the HIF-1α subunit and transcriptional co-activators p300/CBP. Moreover, we demonstrate that KCN1 has significant potential for further development as a therapeutic as it strongly suppresses the growth of malignant glioma cells in vivo without any significant toxicity.

Materials and Methods

KCN1 synthesis and formulation for in vivo delivery

We generated KCN1 (3,4-dimethoxy-N-[(2,2-dimethyl-2H-chromen-6-yl)methyl]-N-phenylbenzenesulfonamide) using a four-step synthesis described in Suppl. Fig. 1A and its structure was confirmed by UV, IR, MS and NMR spectroscopy (not shown). For cell culture experiments, a 10 mM stock solution of KCN1 in DMSO was diluted in pre-warmed media. For animal experiments, we developed a formulation for the delivery of KCN1 by preparing a stock solution (12 mg/ml) in a 1:1 mix of 200 proof ethanol and Cremophor EL (Sigma Cat # C5135-500G) by vortexing and heating to 80-90°C in a water bath. Prior to intraperitoneal (i.p.) administration, the KCN1 stock solution was diluted 1:5 with sterile PBS to a final concentration of 2.4 mg/ml and rapidly administered to avoid precipitation.

Cell culture

The human glioblastoma cell lines (LN229, U251MG, D54MG, D645MG, and LN443) and their growth conditions were previously described (27). The cells were routinely tested for mycoplasma, but no genetic authentication was performed. LN229HRE-luc/lacZ cells were generated by stably transfecting LN229 cells with a bi-directional reporter construct (pBIGL-V6R) in which the firefly luciferase and LacZ reporter genes are under the control of six head to tail tandem copies of the vascular endothelial growth factor (VEGF) HRE in rightward orientation (clone LN229V6R#18; Hygro selection 600 μg/ml) (22). LN229CMV-luc and LN229CMV-lacZ cells were made by stably transfecting CMV promoter-luciferase (CMV-pGL2basic) or CMV promoter-β-galactosidase (CMV-pLacZ) reporters in LN229 cells using G418 (600 μg/ml) selection. Stably transfected cells were maintained in media with appropriate selection agents. Cells were pre-treated with KCN1 or 1% DMSO (final concentration in media) vehicle control for 1h under normoxia (21% O₂); then incubation continued under normoxia or hypoxia (1% O₂) using a hypoxia incubator (Thermo Forma model 3130). KCN1 exhibited significant chemical stability, since HPLC analysis confirmed that it is structurally intact when incubated at 37°C for up to 27 h in cell culture (Suppl. Fig. 2).

Plasmids and transient transfection assays

The VEGF promoter construct was prepared by cloning the [−1181; +95] VEGF gene fragment (numbers are relative to transcription start site, Accession No. AF 095785) in the pGL2basic vector (Promega). The [−173; +31] carbonic anhydrase 9 (CA9) promoter constructs were described previously (28). The CA9 and VEGF HRE mutant promoter constructs contain the ATGCACGTA to ATGCTTTTA [−11; −3] (28) and TACGTGGGC to TAAAAGGGC [−975; −967] mutations, respectively. pNF-κB-luc and pAP1-luc vectors were from Stratagene, p53-responsive PG13PyLuc was described earlier (29). Promoter constructs or the pBIGL-V6R construct were co-transfected with the Renilla luciferase-expressing pRL-CMV (internal control for transfection efficiency) using the Effectene Transfection Reagent (QIAGEN). Cells were exposed to the transfection mixture for 16 h, trypsinized, plated at 40,000 cells/cm², and allowed to adhere for 5 h. The cells were then pre-treated with 10 μM KCN1 or 1% DMSO (final concentration in medium) for 1 h and exposed to normoxia or hypoxia for 24 h in the presence of inhibitor.

Reporter assays

Firefly and Renilla luciferase activities were measured with a Dual-Luciferase Reporter Assay System (Promega) in a 20/20ⁿ Luminometer (Promega). Promoter activities were expressed as the average ratios of firefly to Renilla luciferase activities (±S.D.) from at least three independent experiments performed in triplicates. Firefly luciferase activity in lysates of LN229HRE-luc/LacZ xenograft sections was similarly measured and normalized against protein concentration. For β-galactosidase staining, cells were fixed in 1×PBS containing 0.5% glutaraldehyde, 1.25 mM EGTA, and 2 mM MgCl₂ for 5 min, washed three times with 0.1 M sodium phosphate, pH 7.3, 2 mM MgCl₂, 0.01% deoxycholate, and 0.02% NP-40 for 5 min, stained in washing buffer supplemented with 1 mg/ml X-gal, 5 mM potassium ferrocyanide, and 5 mM potassium ferricyanide for 4 h, and stored in washing buffer.

Northern blot analysis

Total cellular RNA was extracted in Trizol (Fisher), separated by electrophoresis in 1% agarose-formaldehyde gels, and transferred to a nylon membrane (GE Healthcare). VEGF, Glut1, and β-actin mRNA levels were analyzed as previously described (22).

RT-PCR analysis

Total RNA was isolated with an RNeasy mini kit (QIAGEN) and cDNA, synthesized with a ProtoScript first-strand cDNA synthesis kit (New England Biolabs) , was amplified with Angiopoietin-like 4 (Angptl4) (GCCTATAGCCTGCAGCTCAC sense and AGTACTGGCCGTTGAGGTTG antisense), CA9 (CTGTCACTGCTGCTTCTGAT sense and TCCTCTCCAGGTAGATCCTC antisense), VEGF (CCTTGCTGCTCTACCTCCAC sense and CACACAGGATGGCTTGAAGA antisense), and β-actin (ACAACGGCTCCGGCATGTGCAA sense and CGGTTGGCCTTGGGGTTCAG antisense) primer pairs as described previously (28).

VEGF ELISA

VEGF concentrations in cell media or lysates of frozen tumor sections were determined with an ELISA kit (R&D Systems) as recommended and normalized against protein concentrations.

Western blot analysis

Total cell lysates from control and KCN1-treated LN229 cells were separated by SDS-PAGE and probed with anti–HIF-1α (1:600; BD Bioscience), anti–HIF-1β (1:1,000; BD Bioscience), anti–CAIX (1:1,000; Novus Biologicals), anti–β-actin (1:1,000; Santa Cruz Biotechnologies), and anti–histone H1 (1:1,000; Santa Cruz Biotechnologies) antibodies as previously described (22).

Pull-down of the HIF complex with an HRE oligonucleotide

Nuclear extracts, prepared with the NE-PER kit (Pierce), were incubated with a 5′-biotinylated double-strand oligonucleotide comprising two VEGF HRE motifs (underlined) (5′-CCACAGTGATACGTGGGCTCCAACAGGTCCTCTTCCACAGTGATACGTGGGC TCCAACAGGTCCTCTT-3′) in a buffer A (10 mM Tris, pH 8.0, 150 mM NaCl, 12% glycerol, 1 mM DTT) at room temperature for 20 min. Pre-washed streptavidin agarose beads (Pierce) were then added to the samples, incubated with agitation overnight at 4°C, recovered by centrifugation and washed four times with buffer A. Pulled down proteins were detached from the beads by denaturation in a SDS sample buffer (50 mM Tris, pH 6.8, 2% SDS, 10% glycerol, 0.02% bromophenol blue, 1 mM DTT), separated by SDS-PAGE, and HIF-1α and p300 proteins were detected by immunoblotting.

Co-immunoprecipitations (Co-IP)

Nuclear extracts were prepared from cells recovered by scraping in ice-cold PBS and aliquots (0.5 mg) were incubated with 2 μg of a primary antibody and protein G sepharose beads (Amersham) at 4°C overnight. Beads were recovered by centrifugation, extensively washed with 50 mM Tris, pH 8.0, 150 mM NaCl, 1% NP40, 1 mM DTT, and protease inhibitors (EDTA-free Complete, Roche). Samples were denatured in the SDS sample buffer, separated by SDS-PAGE, and antibodies against HIF-1α, HIF-1β, p300 (1:1,000; Santa Cruz), CBP (1:1,000; Abcam), and histone H1 proteins were used for immunoblotting.

Chromatin immunoprecipitation (ChIP) assay

ChIP assays were performed with the EZChIP Assay Kit (Cell Signaling Technology). A total of 4×10⁷ cells were pre-treated with 1% DMSO (control) or 10 μM KCN1 for 1 h, transferred to hypoxia for 24 h, and fixed in 1% formaldehyde at room temperature for 20 min. Isolated nuclei were lysed, followed by chromatin digestion with micrococcal nuclease. Chromatin fragments were immunoprecipitated with anti-HIF-1α, anti-p300, or anti-CBP polyclonal antibodies as mentioned above or rabbit IgG as a control. After reversal of cross-linking and DNA purification, DNA from input (1:20 dilution) or immunoprecipitated samples was analyzed with PCR, and products were separated by 2% agarose gel electrophoresis. Primers used to detect endogenous CA9 HRE were 5′-GACTTTGGCTCCATCTCTGC-3′ (sense) and 5′-GACAGCAGCAGTTGCCACAGT-3′ (antisense).

Animal experiments

In two independent experiments, LN229HRE-luc/lacZcells (5×10⁶) were injected subcutaneously (s.c.) into the flanks of nu/nu mice (athymic nude mice, 2 dorsal injection sites/mouse) as described (30, 31) and the mice were randomized into vehicle control and KCN1 chronic treatment groups (8 mice/group in the first experiment and 15 mice/group in the repeat experiment). The mice received tattoos for easy identification (32). KCN1, prepared daily in ethanol/Cremophor EL (1:1) formulation, was administered at 60 mg/kg i.p. 5 days per week for the duration of the experiment, starting 24 h after tumor cell injection. Tumor volumes were measured weekly as described (32) and animals were sacrificed as per IACUC guidelines and the tumors harvested. Sections of frozen tumors, prepared with a cryostat (Leica CM 1850), were mounted on glass slides, and lysed in 100 μl of 1×Passive Lysis Buffer (Promega) for measurement of luciferase activity and VEGF protein levels. For the evaluation of acute KCN1 effects on hypoxia-inducible genes by RT-PCR, tumors were pre-established for 10 weeks to a size of ~300 mm³, when a single dose of KCN1 was administered. Animals were sacrificed 12 h later and tumors harvested and frozen for subsequent RNA analysis.

For the intracranial experiments, nu/nu athymic nude and C57BL/6 mice were stereotactically inoculated with 1×10⁶ LN229HRE-luc/lacZ or CMV-luc and 5×10⁴ B16LS9 cells, respectively, as described previously (31). KCN1 treatment started on day 10 (LN229 cells) or day 1 (B16LS9 cells) as described above. For the survival analysis, the Kaplan Meier method was used to generate survival curves. The logrank test was used to test the difference in the survival times of different groups.

Quantitative and Statistical analysis

Densitometric scanning was performed with Image J software. Data were analyzed using two-sample T-test (MS Office Excel 2007). Statistical comparisons of in vivo s.c. tumor growth were performed by XenoCat_1.0_3 R package using a categorizing mixed-effects model (33). Differences between control and drug-treated tumor volumes were considered significant when p values < 0.05. * - p value <0.05; ** - p value <0.01; ***- p value <0.001.

Results

KCN1 inhibits HIF-inducible gene expression

Following the screening of a combinatorial library for inhibitors of HIF transcriptional activity, we identified KCN1 (Fig. 1A) as one of the most potent compounds (26, 34-36) KCN1 inhibited hypoxia-induced expression of luciferase in human glioma cells (LN229HRE-luc/lacZ cells) stably transfected with a hypoxia-inducible dual luciferase/β-galactosidase reporter pBIGL-V6R (22) in a dose-dependent manner, with an IC₅₀ of ~593 nM (+/− 63), whereas it had little effect on constitutive luciferase activity in LN229CMV-luc cells (Fig. 1B). Similarly, KCN1 inhibited hypoxia-induced β-galactosidase expression in LN229HRE-luc/lacZ cells but not the constitutive expression in control LN229CMV-lacZ cells (Fig. 1B). KCN1’s ability to inhibit HRE-dependent reporter activity was observed in a genetically diverse set of glioblastoma cell lines (U251MG, D54MG, D645MG, and LN443) as shown by transient transfection with the pBIGL-V6R reporter plasmid (37) (Fig. 1C), indicating that KCN1 is a general inhibitor of HIF transcription in gliomas. KCN1 was not a broad inhibitor of transcription, as reporters for NFκB (five response elements-RE), AP1 (six RE) and p53 (13 RE) transcription factors were not affected (Fig. 1D). KCN1 did not affect cell viability at concentrations below 100 μM in LN229 cells (35, 36) further ruling out the possibility that the inhibitory effect on HRE-driven expression was due to non-specific cytotoxicity. In vitro testing in the NCI-60 Tumor Cell Line Screen confirmed that at 10 μM KCN1 did not significantly inhibit the growth of most cell lines, except for non-small cell lung cancer, melanoma, and leukemia cell lines (Suppl. Fig. 3).

Figure 1. KCN1 specifically inhibits hypoxia-inducible gene expression driven by exogenous HRE-reporter constructs in glioma cells.

A) Chemical formula of KCN1 and IC₅₀ established in a cell-based assay using LN229HRE-luc/LacZ cells with a stably integrated hypoxia-responsive luciferase reporter. N=21 independent experiments performed in triplicates.

B) Left panel: KCN1 inhibits hypoxia (1% O₂)-induced luciferase activity in LN229HRE-luc/lacZ (HRE-luc) but not in LN229CMV-luc (CMV-luc) cells. Data from three independent experiments (N=3) performed in triplicates are expressed as percent (%) of the control activity (1% DMSO) (±S.D.). ** - p value <0.01. Right panel: KCN1 inhibits hypoxia (1% O₂)-induced β-galactosidase activity in LN229HRE-luc/lacZ but not in LN229CMV-lac Z cells. β-galactosidase activity was detected by chemical staining and positive cells were quantified.

C) KCN1 inhibits hypoxia-inducible gene expression in other glioma cell lines. Cell lines were transiently co-transfected with pBIGL-V6R-HRE-luc construct and pRL-CMV and tested under normoxia or hypoxia (1% O₂) +/− KCN1 (10 μM). Average of the ratio of luciferase and Renilla activities (± S.D.) from three independent experiments (N=3) performed in triplicates was calculated and promoter activities are expressed as fold induction over the normoxic control set as 1. ** - p-value < 0.01.

D) KCN1 (10 μM) inhibits the activity of a HIF-activated luciferase reporter construct, but not that of reporters for other transcription factors. Constructs were transiently transfected with pRL-CMV in LN229 cells and tested under normoxia or hypoxia (1% O₂). Data from three independent experiments (N=3) performed in triplicates are expressed as percents (%) of the vehicle treated controls (1% DMSO) (±S.D.). ** - p value <0.01.

Next, we tested whether KCN1 inhibits the endogenous target genes of HIF-1. First, we probed levels of VEGF, Glut1 and CA9 mRNAs in glioma cells by Northern blotting and RT/PCR. Compared to the vehicle controls, KCN1 significantly reduced hypoxia-induced levels of these transcripts, whereas it had no effect on β-actin mRNA (Fig. 2A). VEGF ELISA revealed that hypoxia potently activates the secretion of VEGF by LN229 cells and this was antagonized by KCN1 (Fig. 2B, left panel). Transcriptional activity of a wt (but not HRE mutant) VEGF promoter-driven reporter was also significantly upregulated by hypoxia and inhibited by KCN1 (Fig. 2B, right panel). Immunoblotting confirmed that expression of CAIX was also down-regulated by KCN1 (Fig. 2C, left panel). Hypoxia-activated expression from the reporter construct containing the wt CA9 promoter was also inhibited, albeit to a lesser extent than the one driven by the VEGF promoter (Fig. 2C, right panel). Acute KCN1 treatment also inhibited expression of endogenous hypoxia-inducible genes in vivo in pre-established xenografts of LN229V6R cells (Fig. 2D). Together, these data show that KCN1 inhibits expression of representative endogenous hypoxia-inducible genes that encode regulators of important tumor functions such as angiogenesis, glucose transport and cellular pH.

Figure 2. KCN1 inhibits the expression of endogenous hypoxia-inducible genes.

A) Left panel: Northern blot analysis of VEGF and Glut-1 gene expression in LN229 cells under normoxia and hypoxia (1% O₂) +/− KCN1 (25 μM) treatment for 4 to 24 hours. β-actin was used as a loading control. Right panel: Densitometric analysis of the Northern blot. Data are expressed as the ratio of VEGF(Glut1)/β-actin signal. Bottom panel-left: RT-PCR analysis of CA9 and β-actin gene expression with and without KCN1 (25 μM) treatment in U251MG cells. Right: Relative CA9 expression levels were quantified by densitometry and normalized against β-actin expression (average from three different gels ± S.D.).

B) Left panel: ELISA-mediated detection of VEGF levels in the 48h conditioned media of LN229 cells grown under normoxia (21% O₂) or hypoxia (1% O₂) +/− KCN1 (10 μM) treatment (left panel). Right panel: Activity of VEGF promoter constructs with wild type (wt) or mutant HRE in transiently transfected LN229 cells +/− KCN1 (10 μM) (right panel). A constitutively active CMV-Renilla luciferase construct was co-transfected as a control. Average of the ratio of luciferase and Renilla activities (± S.D.) from three independent experiments (N=3) performed in triplicates was calculated and promoter activities are expressed as fold induction over the normoxic control set as 1.

C) Left panel: Western blot mediated detection of CAIX expression in LN229 cells under normoxia or hypoxia +/− KCN1 (10 μM) (left panel). Akt expression was used as a loading control. Relative CAIX expression levels were quantified by densitometry and normalized against Akt expression (average from two different blots ± S.D.). Right panel: Activity of CA9 promoter constructs with wild type (wt) or mutant HRE in transiently transfected LN229 cells +/− KCN1 (10 μM) (right panel). Normalization to the CMV-Renilla luciferase reporter was done as in B. ** - p value <0.01; * - p value <0.05. D) Relative Angptl4, CA9, and VEGF mRNA expression levels in control (vehicle) and KCN1-treated LN229V6R tumors analyzed by RT-PCR. Mice with pre-established LN229 tumors received a single dose of KCN1 (60mg/kg i.p.) and the tumors were excised 12h later. Intensity of each Angptl4, CA9, VEGF, and β-actin band in agarose electrophoresis was quantified by densitometry and expression of hypoxia-inducible genes was normalized against β-actin expression (average from three different experiments ± S.D.). Results on tumors from two vehicle-treated (C5L, C5R) and two KCN1-treated (K5L and K5R) animals are shown. P-values for Angptl4, CA9, and VEGF expression in control and KCN1-treated tumors were 0.000145, 0.00263, and 0.03042, respectively.

KCN1 is anti-tumorigenic in vivo

To assess the anti-tumor potential of KCN1, we examined its anti-glioma activity in in vivo animal models. LN229HRE-luc/lacZ cells were inoculated s.c. into both flanks of nu/nu mice and treated i.p. with KCN1 (60 mg/kg) or vehicle 5× per week. Cumulative data from two independent experiments show that after a 10-week chronic treatment period the average tumor volume in KCN1-treated animals was >4-fold lower than that in the controls (Fig. 3A). Remarkably, 10 out of 42 tumors in the KCN1 group grew till about 20-60 mm³ after 4-6 weeks, then completely regressed. Six more tumors showed stable disease at tumor sizes of 20-170 mm³. When grouped, the dissected tumors in both experiments revealed a ~3-fold weight difference between the groups (Fig. 3B, C). Finally, the average tumor growth rates measured were also found to be significantly lower in the KCN1-treated group compared to the control group (Fig. 3C). KCN1 was further tested in a s.c. model where tumors were allowed to form for 3.5 weeks (~26 mm³) before the treatment started. Tumor growth curves in Suppl. Fig. 4 indicate that KCN1 shows significant antitumor activity in pre-established tumors, although it was most potent as a prevention agent. KCN1 was also anti-tumorigenic in animal models of pancreatic cancer (38) and uveal melanoma and its metastasis to the liver (70% reduction of tumor size in the eye and 50% reduction in number of hepatic metastases; manuscript in preparation). KCN1 injections were well tolerated and did not evidence any signs of extraneous toxicity; the animals’ behavior and activity were indistinguishable from untreated animals, and their bodily appearance and weight were normal (Suppl. Fig. 5). Initial pathological examination of main organs demonstrated no ultra-structural changes in brain, kidney, GI tract and lung (Suppl. Fig. 6). A treatment-related change was observed in the liver, where swelling was evidenced macroscopically at autopsy, and pathology showed tissue edema with bile duct stasis, yet without evidence for any hepatocyte death. The swelling was reversible within 2-3 weeks after treatment was discontinued (data not shown) and may be related to the cremophor:ethanol formulation, which can interfere with hepatic blood flow (39, 40).

Figure 3. KCN1 inhibits the growth of malignant human glioblastoma xenografts, HRE activity and VEGF levels in vivo.

Nude mice carrying LN229HRE-Luc/LacZ cells (s.c. in the flanks; 2 tumors/mouse) were treated i.p. 5 days per week with 60 mg/kg KCN1 in 1:1 ethanol/cremophor (controls vehicle only).

A) Tumor volume. The growth curve shown is the combined result of two independent experiments. The first was conducted with 8 mice/group and the second with 15 mice/group. Average tumor size in each group (+S.E.) is shown. p<0.001.

B) Size (left) and mass (right) of tumors excised from the first experiment (2 tumors/mouse). ** p<0.01.

C) Left panel: Individual tumor sizes of both experiments combined. Right panel:Kinetics of tumor growth from week 9 to 10 expressed as increase in tumor size per day. *** p< 0.001.

D) Left panel: Effect of KCN1 on HRE-luciferase activity in tumors after 10 weeks of KCN1 treatment. Luciferase activity was measured in lysates of frozen tumor sections and normalized against protein concentration. * p= 0.017. Right panel: VEGF levels in tumors after 10 weeks of KCN1 treatment. VEGF protein in lysates of frozen tumor sections was detected by ELISA and normalized against protein concentration. * p= 0.014.

To gain an insight whether the tumors that did not regress after the 10-week KCN1 treatment period might have become resistant to the anti-HIF activity of KCN1, we analyzed the tumors for production of luciferase from the HIF-inducible reporter and endogenous VEGF in the drug- and vehicle-treated groups. Although tumor-to-tumor variation in luciferase activity and VEGF levels in each group was observed, the average values in both cases were still significantly lower (p value <0.05) in KCN1-treated tumors (Fig. 3D), confirming that in these resilient tumors KCN1 is still inhibiting HIF activity and VEGF production to some extent.

We then tested whether KCN1 would show efficacy in orthotopic brain tumors. We used two intracranial tumor models: one with LN229 human glioma cells in nu/nu athymic nude mice and the other with B16LS9 mouse melanoma cells in syngeneic C57BL/6 mice. KCN1-treatment temporarily reduced mortality rate in mice injected with LN229 cells (significance for days 41-48 was p<0.05 and for days 42-47 p<0.01), although the survival end-point showed no difference between the groups (Suppl. Fig. 7A and B). No survival benefit was observed in the more aggressive B16LS9 melanoma model (Suppl. Fig. 7C).

We further determined whether KCN1 showed blood brain barrier (BBB) permeability, using in vitro BBB assays with 14C*-labeled KCN1 (41). Measuring of apical to basolateral brain endothelial cell permeability revealed that KCN1 has a permeability coefficient comparable to mannitol, a cell-impermeable control (Suppl. Fig. 8). Our pharmacokinetic studies found that the concentration of KCN1 in brains of CD1 mice (~0.4 μM) was at least 5x lower than in other organs (spleen, kidney, liver, lung) after i.p. and intravenous administration (38). Together, these data are indicative of the absence of an active influx mechanism for KCN-1, suggesting that it may not penetrate an intact BBB efficiently.

Collectively, these findings provide proof-of-principle for the anti-tumor efficacy, inhibition of HIF activity in vivo, and low toxicity of KCN1 and provide the basis for further pre-clinical development of this class of agents. Chemical modifications to increase their potency, optimize their pharmacology, and enable their BBB permeability are still warranted.

KCN1 does not decrease levels of components of the HIF-1 complex

To characterize the molecular mechanism of inhibition of HIF-1-dependent transcription, we initially examined the effect of KCN1 on levels of components of the HIF-1 complex. Western blot analysis of total cell lysates from LN229 cells revealed that over the 12-48 h treatment period the HIF-1α protein levels showed little variation in response to KCN1 (Fig. 4A). Other constituents of the HIF-1 complex (HIF-1β and p300) were also unaffected by KCN1 treatment. Similar results were obtained with other cell lines (U87MG, HEK293ft) in which KCN1 had been shown to inhibit HRE activity (data not shown). These observations establish that KCN1 does not inhibit HIF activity by down-regulating the levels of components of the HIF-1 transcriptional complex.

Figure 4. The mechanism of action of KCN1 involves the disruption of the HIF-1α, HIF-1β, p300/CBP transcription complex.

A) Western blot of whole cell lysates shows that KCN1 (25 μM) treatment does not significantly alter the levels of the HIF-1α, HIF-1β subunits or of the p300 transcription co-factor in hypoxic (1% O₂) LN229 cells over a 12-48 h time frame.

B) Pull-down experiments demonstrate that KCN1 reduces the interaction between p300 and the HRE sequence. Top panel: A biotinylated HRE probe was incubated with LN229 nuclear extract, complexes were pulled-down with streptavidin-agarose beads and tested for the presence of HIF-1α and p300 by Western blotting. Bottom panel: Densitometric analysis of the p300 pull down. Data are expressed as the ratio of p300 pull down signal/ p300 input signal.

C) Chromatin immunoprecipitation (ChIP) analysis demonstrates a reduction in the binding of p300 and CBP co-factors on the CA9 HRE by KCN1 (10 μM) in LN229 cells, while that of HIF-1α was unaffected.

KCN1 prevents the binding of transcriptional co-factors p300 and CBP on the HRE sequence in vitro and in vivo

We next interrogated how KCN1 affects the assembly of a functional HIF-1 transcriptional complex on an HRE sequence using a DNA-protein in vitro pull-down experiment. A biotinylated double-strand oligonucleotide with two copies of the VEGF HRE sequence was exposed to nuclear extract from hypoxic glioma cells and pulled-down complexes analyzed by immunoblotting for the presence of p300 and HIF-1α (Fig. 4B). The input fractions confirmed that KCN1 does not affect HIF-1α and p300 levels and analysis of the pulled down fractions showed that the HRE binding activity of HIF-1α is not compromised by KCN1. In contrast, KCN1 decreased the amount of HRE-bound p300 in a dose-dependent manner, hinting that KCN1 may prevent the recruitment of the p300 cofactor to the pre-assembled HRE-HIF-1 complex. To further establish whether KCN1 prevents the recruitment of p300 and CBP to the chromatin of an endogenous hypoxia-inducible gene in vivo, we studied the assembly of the individual protein components of the HIF complex on the CA9 gene HRE by chromatin immunoprecipitation (ChIP) assay. ChIP data indicated that KCN1 had no effect on the binding of HIF-1α to the CA9 HRE DNA sequence, corroborating the conclusion from the pull-down experiment that KCN1 does not affect HRE-binding activity of HIF-1. Consistent with the pull-down assays, the levels of p300 and CBP were decreased in the complex assembled on the CA9 HRE of KCN1-treated cells under hypoxia (Fig. 4C).

KCN1 interferes with the binding of HIF-1α to the transcriptional co-activators p300 and CBP

To more directly examine how KCN1 inhibits HIF-1α-p300/CBP assembly on the HRE, we performed co-immunoprecipitation (co-IP) experiments. Both the co-IP of p300 with an HIF-1α antibody and the reversed co-IP of HIF-1α with a p300 antibody demonstrated that KCN1 interferes with the HIF-1α-p300 interaction (Fig. 5A, B). Similarly, a co-IP with an antibody against the p300 paralog, CBP, also showed that the HIF-1α-CBP interaction is negatively affected by KCN1 (Fig. 5C). On the other hand, KCN1 had no appreciable effect on the HIF-1α-HIF-1β heterodimer formation (Fig. 5D).

Figure 5. KCN1 disrupts the HIF-1α, HIF-1β, p300/CBP transcription complex.

Co-immunoprecipitations (Co-IP) in cell extracts of LN229 cells pretreated with KCN1 (25 μM) show a reduction in binding between HIF-1α and p300/CBP co-factors. A Western blot on the cell extract was performed as a control for equal protein distribution (Input). H-1, Histone H-1.

A) Co-IP of p300 with HIF-1α.

B) Co-IP of HIF-1α with p300.

C) Co-IP of HIF-1α with CBP.

D) Co-IP of HIF-1β with HIF-1α.

Combined, the ChIP, HRE pull-down and co-IP assays support the notion that the mechanism of inhibition of HIF activity by KCN1 involves interference with HIF-1α-p300/CBP interaction, which causes p300/CBP deficiency in the HIF complex.

Discussion

A growing body of evidence supports the facilitating role of HIF, and HIF-regulated gene products, in tumor progression/metastasis; therefore, HIF is increasingly considered an attractive therapeutic target. The rationale for targeting HIF in cancer is that HIF is the key transcription factor responsible for the transactivation of a wide array of genes, many of which enhance survival and spread of tumor cells (13, 14). Blocking of HIF function would thus be expected to interfere with multiple attributes of tumor cells and eventually lead to tumor regression. Not surprisingly, significant effort and resources have been invested into identifying small molecules that would potently and specifically inhibit HIF.

Our laboratory has had a long-standing interest in the HIF pathway and the development of novel experimental therapeutics targeting the hypoxic status of tumors (22, 23, 37, 42-46). To identify new inhibitors of the HIF pathway, we have developed a cell-based assay as this format is more likely to yield pharmaceutically viable leads (47). The screen of a combinatorial library of natural product-like compounds (25) yielded a novel class of agents (arylsulfonamides) that inhibited HIF activity at high nano to low micromolar concentrations (26). Here we showed that the initial lead compound in this class, KCN1, displayed inhibition of HIF-dependent expression in the context of artificial HRE enhancers as well as HREs in the endogenous HIF-inducible genes. Inhibition was dependent on the presence of a functional HRE, as evidenced by differential effects of KCN1 on wt and HRE mutant VEGF and CA9 promoter constructs. More importantly, KCN1 displayed significant anti-tumor activity in vivo. Systemic administration of KCN1 to mice harboring s.c. malignant gliomas markedly inhibited tumor growth and HIF activity in tumors in the absence of any significant toxicities even when the compound was administered daily at 60 mg/kg for extended periods of time. While some liver swelling was noticed at autopsy, this was reversible and likely linked to the use of Cremophor:ethanol (39, 40), suggesting that a new formulation or more soluble analogs needs to be developed, a process we recently started with the synthesis of heteroarylsulfonamides (35, 36)

In contrast to the anti-tumor efficacy seen in s.c. cancer models, KCN1 showed little therapeutic effect when tested in orthotopic brain tumor models (glioma and melanoma). Pharmacological assessment of KCN1 distribution to the brain tumors and in vitro BBB permeability assays suggest that new analogs need to be developed with improved BBB permeability. Regulation of HIF-1 levels and activity is a complex multi-step process, controlled primarily by HIF-specific factors, PHDs and FIH. In addition, HIF is situated at the convergence of major oncogenic signaling pathways (PI3-K, MAPK, mTOR) that activate the HIF pathway non-specifically. Not surprisingly, the complexity of HIF regulation (transcription, translation, folding, transport, proteasomal degradation of HIF-α, DNA binding, and interaction with transcriptional co-activators) provides multiple steps that are “druggable” and small molecule inhibitors intervening at various stages of the regulatory process have been identified (15, 20, 47, 48). A large number of agents initially developed to target signaling pathways, such as the inhibitors of PI3K, mTOR, MAPK, topoisomerase II, and even the modulators of microtubule dynamics, have been also shown to indirectly inhibit HIF-1 function (48). The search for new, more specific HIF inhibitors has already provided a number of novel agents: acriflavine, a specific inhibitor of HIF-1α/HIF-1β dimerization (16); and echinomycin and “programmable” polyamides that disrupt the interaction of HIF-1 with the HRE through DNA intercalation (49). Several of these agents are being translated into clinical trials, whereas others have shown unacceptable toxicity. The difficulty in identifying compounds targeting the HIF-1α–p300/CBP interaction is underscored by a prior extensive chemical screen using a large library of 600,000 small molecules, which identified a single compound, chetomin, which displayed anti-tumor activity in vivo, albeit in the presence of some toxicity (12). Later work suggested that, rather than specifically targeting p300, chetomin chelates Zn²⁺ that is required for the structure and function of its CH1-3 domains (50).

In a cell-based assay the molecular target(s) of an inhibitor is not immediately identified and its identification can become a major challenge. Many of the prior reported HIF inhibitors, including 103D5R and KC7F2 we previously isolated (21, 22), suppress HIF function by reducing HIF-1α subunit levels. In contrast, we found that KCN1 does not appreciably alter the levels of HIF-1α, HIF-1β, or p300, suggesting that it may either affect the assembly or function of the HIF complex. We demonstrated, using pull-downs, co-IPs, and ChIP, that KCN1 compromises interactions between HIF-1α and transcriptional co-activators p300/CBP, which impairs the recruitment of these co-factors to preassembled HRE-HIF complexes on the chromatin, and prevents hypoxia-induced transcription. The interaction between HIF-1α and p300/CBP is mediated by the CAD of HIF-1α and the CH1 domain of p300/CBP (10) and to a lesser extent by the N-terminal activation domain of HIF-1α with the CH3 domain of p300/CBP (11) Our recent modeling studies evidenced putative binding sites for KCN1 on the CH1 domains of p300 and CBP, which are predicted to block the interaction with HIF-1α (41).

In summary, we have identified and characterized a novel type of HIF pathway inhibitor – a di-substituted sulfonamide with the naturally occurring 2,2-dimethyl-2H-chromene structural motif, designated KCN1. KCN1 inhibits HIF activity in an HRE-dependent cell-based assay with an IC₅₀ of ~590 nM, and displays promising anti-tumor activity in animal models. KCN1 inhibits HIF function in a unique way, with the mechanism of action involving interference with HIF-1α-p300/CBP interactions (first-in-class), which results in co-activator deficiency in the HIF complex assembled on the target genes under hypoxia and in turn decreases transcription activity. HIF inhibitory activity may not be the only factor contributing to the antitumor effect of KCN1 and there could be other relevant targets. Further studies are warranted for translation of this promising chemotype toward clinical development, which include our ongoing structure-activity relationship studies directed towards improvement of potency and pharmacological properties.

Translational relevance.

Glioblastoma (GBM), one of the most aggressive and lethal cancers with a life expectancy of less than one year, currently has no effective cure. The hypoxia inducible factor-1 ( H I F-1) plays a critical role in tumor adaptation to the hypoxic microenvironment and associates with poor prognosis and treatment failure. The HIF pathway plays a prominent role in GBM and targeting the HIF pathway has become an important therapeutic strategy. Here, we evaluate for the first time the anticancer activity of arylsulfonamide KCN1, the lead inhibitor in a novel class of small molecule inhibitors of the HIF-1 pathway we recently discovered. Our results show that KCN1 displays potent anti-glioma activity in a subcutaneous mouse model, whereas its activity was more limited in orthotopic brain tumor models. Arylsulfonamides thus have therapeutic potential for GBM and other human tumors although further optimization of their pharmacological properties will be required.

Abbreviations

Angptl4

Angiopoietin-like 4

CA9

carbonic anhydrase 9

CAD

C-terminal activation domain

CH1

cysteine-histidine rich 1 domain

ChIP

chromatin immunoprecipitation

Co-IP

Co-immunoprecipitation

FIH-1

Factor Inhibiting HIF-1

HIF-1

hypoxia inducible factor-1

HRE

hypoxia-responsive element

i.p.

intraperitoneal

RE

response element

s.c.

subcutaneous

VEGF

vascular endothelial growth factor