Benzylidene‐indolinones are effective as multi‐targeted kinase inhibitor therapeutics against hepatocellular carcinoma

Abstract

Effective pharmacological intervention of advanced hepatocellular carcinoma (HCC) is currently lacking. Despite the use of tyrosine kinase inhibitors (TKIs) for the targeted therapy of several malignancies, no agent has been developed to specifically interfere with the oncogenic tyrosine kinase signaling aberrations found in HCC. Therefore, we adopted an orthogonal biological phenotypic screening approach to uncover candidate compounds: based on a potent cytotoxicity toward HCC‐derived cell lines, and minimal toxicity toward normal liver cells. Given the success of indolinone as a chemical scaffold in deriving potent multi‐kinase inhibitors (e.g. sunitinib), we screened a group of newly synthesized benzylidene‐indolinones. Among the candidates, E/Z 6‐Chloro‐3‐(3‐trifluoromethyl‐benzyliden)‐1,3‐dihydroindol‐2‐one (compound 47) exhibited potent anti‐proliferative, anti‐migratory, pro‐apoptotic properties and good safety profile as compared to known multi‐targeted tyrosine kinase inhibitors sunitinib and sorafenib. Additionally, an accompanying suppression of alpha‐fetoprotein (AFP) transcription, an HCC tumor marker, implies a favorable selectivity and efficacy on HCC. The in vivo efficacy was demonstrated in an HCC xenograft where 47 was administered once weekly (60 mg/kg) and suppressed tumor burden to the same extent as sorafenib (30 mg/kg daily). A receptor tyrosine kinase (RTK) array study revealed promising inhibition of multiple tyrosine kinases such as IGF‐1R, Tyro3 and EphA2 phosphorylation. Gene silencing of these targets ameliorated the cytotoxic potential of 47 on the HuH7 cell line, thereby implicating their contribution to the tumorigenicity of HCC. Hence, 47 exhibits potent anti‐cancer effects on HCC cell lines, and is a suitable lead for developing multi‐targeted kinase inhibitors of relevance to HCC.

Highlights

• A phenotype‐based screening of novel indolinones uncovers potent agents against HCC.

• Design helps identify effective compounds with minimal effect on healthy hepatocytes.

• 47 Exhibits superior anti‐cancer properties vs. sunitinib, including AFP suppression.

• In vivo efficacy of 47 was demonstrated with orthotopic HCC model.

• Inhibition of multiple tyrosine kinases contribute to the observed effects.

Abbreviations

HCC

hepatocellular carcinoma

TKI

tyrosine kinase inhibitor

RTK

receptor tyrosine kinase

AFP

alpha-fetoprotein

47

Compound 47

1. Introduction

Over the past 15 years, tyrosine kinase inhibitors (TKIs) have emerged as important pharmacological agents in the area of targeted therapy against cancers, including hepatocellular carcinoma (HCC). Conventionally, the mechanistic basis for targeted therapy hinges on monogenic lesions such as the Bcr–Abl fusion and translocation in chronic myelogenous leukemia, or HER2 amplification in breast carcinomas, which provide exclusive signals for driving tumorigenesis and neoplastic progression (Weinstein, 2002). However, mounting evidence suggests that kinase mutations for many malignancies are sporadic rather than congregating on specific “hotspots” in the genome (Bleeker and Bardelli, 2007; Greenman et al., 2007; Ruhe et al., 2007; Thomas et al., 2007b). This implies that the underlying mechanism driving disease progression involves multiple oncogenic defects which would thus undermine the effectiveness of therapy directed against a single target. Moreover, recent studies have shown that receptor cross‐talk is pervasive and provides the means for tumors to switch oncogene dependence. For example, EGFR inhibition led to secondary amplification of MET to sustain the growth of non‐small cell lung carcinoma (Engelman et al., 2007), and IGF‐1R signaling was found to be upregulated to circumvent growth arrest caused by erlotinib (Buck et al., 2008). Thus, targeting multiple pathways may reduce or delay the incidence of oncogenic kinase‐switching and hence the onset of chemoresistance in chronic cancer therapy (Petrelli and Giordano, 2008; Weinstein and Joe, 2008).

HCC exemplifies a condition where a definitive disease‐driving mutation in the cellular signaling network remains elusive. Several tyrosine kinases have been associated with HCC development and growth. For instance, overexpression of FAK, PYK2, IGF‐1R and FGFR3 were reported separately to contribute to disease severity (Itoh et al., 2004; Qiu et al., 2005; Scharf and Braulke, 2003; Sun et al., 2007). More recently, we observed elevation of FGFR4 and its co‐receptor in many HCC patients during normal‐to‐tumor transition and proposed a potential link to AFP regulation (Ho et al., 2009; Poh et al., 2012). In spite of some progress in this area, research in TKIs development for HCC lags behind that of other cancers. Currently, clinical trials of TKIs in HCC are based on approved or investigative drugs that were initially designed for other diseases. Not surprisingly, the empirical nature of these trials has yielded marginal effects because the critical tyrosine kinases underlying the disease may not have been targeted. For instance, phase II clinical studies with imatinib and erlotinib in unresectable HCC yielded disappointing responses, with most patients showing disease progression and none exhibiting complete or partial response (Lin et al., 2008; Thomas et al., 2007a).

Among currently approved TKIs, the “gold standard” against HCC is sorafenib, which was approved by FDA for HCC in 2007 (Simpson and Keating, 2008). Sorafenib targets VEGFRs, PDGFRs, RET, c‐Kit as well as Raf kinase. Pivotal trials reported an unprecedented increase in overall survival (about 3 months) in HCC patients (Abou‐Alfa et al., 2006; Llovet et al., 2008). Sunitinib, another multi‐targeted TKI, has shown promising early results but was limited due to adverse effects and non‐superior benefits over sorafenib (Worns et al., 2010; Zhu, 2008). Several other kinase inhibitors with multi‐targeting effects are in different stages of clinical trials (Chan and Yeo, 2012). However, a recent meta‐analysis has uncovered an increased risk of hepatotoxicity for many of the tyrosine kinase inhibitors with at least 5 TKIs who have received black box warning for liver toxicities (Teo et al., 2013). Inadvertently, this phenomenon can impose an additional burden on the already diseased liver when used in therapy. Therefore, these findings suggest that while a multi‐kinase targeting approach may present new opportunities against HCC, non‐selective liver toxicity will need to be mitigated early in the drug development phase.

Hence, we explored new kinase inhibitor‐like candidates using a dual cell line biological phenotype as an initial screening platform. The basis of selection is to identify potent compounds in blocking the growth HCC cells, while sparing the non‐cancerous liver cells. Benzylidine‐indolinone, a chemical scaffold with good track record of deriving kinase inhibitors was used to discover novel compounds of clinical value. Subsequent in vitro and in vivo assays were then implemented to clarify the mechanism of action and to explore the potential for clinical development. Through this effort, we have found a novel benzylidene‐indolinone E/Z 6‐Chloro‐3‐(3‐trifluoromethyl‐benzyliden)‐1,3‐dihydroindol‐2‐one (47) with a superior safety–efficacy profile against HCC cell lines as compared to sunitinib. This investigation provides the rationale for subsequent lead optimization to improve the drug‐like properties of 47 as a multi‐targeted TKI, and also uncover a panel of tyrosine kinases that might play a novel role in HCC.

2. Experimental procedures

2.1. Chemicals

The series of benzylidene‐indolinones (compounds 39–48) were synthesized using methods described previously (Zhang and Go, 2009). A US patent has been filed by Go ML, Zhang W. Functionalized indolinones, US Patent pending, US Provisional Patent Application No 61/105, 206, pp 26 2008. Sunitinib was obtained from LC Laboratories (Boston, MA). All other reagents were obtained from Sigma–Aldrich (St. Louis, MO) unless otherwise stated.

2.2. Cell culture

HuH7 was obtained from the Max Planck Institute for Biochemistry, Martinsried, Germany. HepG2, SK‐Hep1, Hep3B, PLC/PRF/5, THLE2, Hs817T and HUVEC cells were from ATCC (Manassas, VA). HepG2, Hep3B and PLC/PRF/5 were maintained in MEM, THLE2 in BEGM (Lonza, Basel, Switzerland), and all other cell lines in DMEM. All cell cultures were supplemented with ATCC recommended reagents obtained from Invitrogen (Carlsbad, CA).

2.3. Quantitative real‐time PCR

Total RNA was harvested with Trizol (Invitrogen) as previously described and purified with RNeasy column (Qiagen, Valencia, CA) (Ho et al., 2009). Two μg RNA was used for cDNA synthesis using SuperScript III (Invitrogen), performed according to manufacturer's instruction. Quantitative real‐time‐PCR was performed using ABI 7300 (Foster City, CA) for AFP (NM_001134) with 18S mRNA as control. Samples were prepared in triplicates with 4 μL of pre‐diluted cDNA each. The primer sequences were: AGCTTGGTGGTGGATGAAAC (AFP 5′ forward); TCTTGCTTCATCGTTTGCAG (AFP 3′ reverse); CGGCTTAATTTGACTCAACACG (18S 5′ forward); TTAGCATGCCAGAGTCTCGTTC (18S 3′ reverse). Data were obtained as average C _T values, and normalized against control as ΔC _T. Expression changes in AFP transcripts between normal vs. tumor tissue were expressed as fold change using $2^{\Delta \Delta C_{\text{T}}}$ (difference between the ΔC _T of matched pairs).

2.4. Immunoblot assay

Protein concentrations were assayed by BCA method (Pierce, Rockford, IL). Samples were resolved using 6–10% SDS‐PAGE and transferred to nitrocellulose or PVDF membranes. Immunodetection was by chemiluminescence (SuperSignal, Pierce) using specific antibodies diluted in PBS with 0.05% (v/v) Tween 20 and 5% (w/v) powdered milk or BSA. Anti‐phospho‐IGF‐1R, anti‐phospho‐Erk, anti‐phospho‐Akt, anti‐Bax, anti‐BCL‐xL were from Cell Signaling Technology (Beverly, MA); anti‐PCNA and anti‐HSP60 from Santa Cruz Biotechnology (Santa Cruz, CA); anti‐cyclin‐D1 from BD Pharmingen (San Jose, CA). Secondary antibodies were anti‐mouse and anti‐rabbit HRP‐conjugated antibodies (Pierce).

2.5. Immunoprecipitation

HuH7 cells were treated with 47 or sunitinib (1 or 10 μM) for 2 h or 24 h and harvested for immunoprecipitation using lysis buffer (1% NP‐40, 20 mM Tris–HCl, 137 mM NaCl, 10% glycerol, 2 mM EDTA, 1 mM sodium orthovanadate, 10 μg/mL aprotinin, 10 mg/μL leupeptin). Lysates (300 μg) were diluted in 500 μL of HNTG buffer (250 mM HEPES, 150 mM NaCl, 10% glycerol, 0.1% Triton‐X) and incubated with 30 μL of Protein A/G mix (GE Healthcare, Waukesha, WI) and 2 μg of anti‐EGFR (Upstate, Lake Placid, NY), anti‐EphA2 (Santa Cruz) or anti‐Tyro3 (Bethyl Laboratories, Montgomery, TX) overnight at 4 °C. Resulting beads were washed with HNTG buffer before boiling in 20 μL SDS‐PAGE sample buffer for electrophoresis. Subsequent immunodetection was with anti‐phosphotyrosine (4G10, Upstate) and loading determined by stripping and re‐probing with respective primary antibodies.

2.6. Cell viability assay

Cells were plated at 5000–7000 cells/per well in 96‐well plates and incubated overnight. Various concentrations (0–10 μM) of test compounds (compounds 39–48 and sunitinib) were added to cells for 72 h with 8 replicates. Cell Titer‐Glo (Promega, Madison, WI) assay was performed according to manufacturer's instruction with luminescence detected on SpectraMax M5 (Molecular Devices, Sunnyvale, CA). Data were expressed as percentage of viability vs. vehicle‐treated controls.

2.7. Caspase‐3 activity assay

Cells were subjected to 47 exposure for 24 h prior to harvesting. Caspase‐3 activities were determined by as previously described (Ho et al., 2009). Data were expressed as RFU/μg lysate/h incubation and error bars in terms of SD with n = 3.

2.8. Transwell migration assay

Serum‐starved HuH7 cells were seeded onto 24‐well transwell inserts (BD Biosciences, San Jose, CA) at 200,000 cells per well and allowed to migrate across 8 μm pores using 10% FCS as chemoattractant. Cells were incubated in the presence of 47, sunitinib or DMSO control. At 16 h, cells were scraped from the top and the migrated cells stained with 0.5% crystal violet. Cells were counted in 3 separate fields under 100× magnification. The triplicate of each sample is averaged and data presented as number of cells migrated per well with ±SD (n = 9).

2.9. Phospho‐RTK profiling

Simultaneous determination of multiple RTK phosphorylation was achieved with Human Phospho‐RTK array (RnD Systems, Minneapolis, MN). HuH7 cells were subjected to vehicle, 47 (1, 5 10 μM) or sunitinib (10 μM) treatments for 24 h under serum‐free conditions. Cells harvesting, hybridization with RTK array and incubation with anti‐phosphotyrosine were performed according to manufacturer's instruction. Imaging and quantification of the spots intensities were with Fujifilm LAS‐3000 (Tokyo, Japan). Average signal of the duplicate spots for each RTK was determined.

2.10. Gene silencing by siRNA

Custom‐made ON‐TARGETplus siRNA designed for silencing IGF‐1R, Tyro‐3 and EphA2 were purchased from Dharmacon (Chicago, IL). Four target sequences for each of IGF‐1R, Tyro‐3 and EphA2 were tested for knock‐down efficiency and the best sequences were employed for subsequent caspase‐3 activity assays (see Supplementary Information for siRNA sequences and immunoblots). HuH7 cells were transfected with 50 nM siRNA in 24‐well format, in accordance to manufacturer's instruction. After 72 h incubation with respective siRNAs, cells were re‐seeded onto 96 wells plate. HuH7 cells were subjected to vehicle or 47 treatments for 24 h. These cells were then harvested for both immunoblot and caspase assays performed as described previously.

2.11. Orthotopic HCC model

All procedures involving animals were reviewed and approved by the SingHealth Animal Use and Care Committee. Two million HuH7 tumor cells stably expressing firefly luciferase (HuH7‐Luc2) were implanted subcutaneously at the rear flank of 10 week‐old Balb/c nude female mice (Biolasco, Taiwan). Tumor growth was monitored twice weekly and harvested when tumor volume reached 1–1.5 cm³. After harvesting, the tumor was rinsed with saline and cut into cubes of approximately 1 mm³. Ten week‐old Balb/c nude female mice were implanted orthotopically with the 1 mm³ HuH7‐Luc2 tumor cube. Tumor growth was monitored bi‐weekly by bioluminescence imaging using the IVIS 200 Bioluminescence Imaging System (Xenogen Corp., Alameda, CA). Mice were randomized and treatment was started once the observed bioluminescent signal increased steadily. Mice are randomized into three groups (n ≥ 6). Mice were treated with either 47 at 60 mg/kg or solvent alone via intraperitoneal (ip) injection once weekly. For comparison of efficacy, sorafenib was given to the third group of mice at 30 mg/kg via oral gavage once daily. Tumor growth and body weight were monitored every 3–4 days. For imaging, mice were given ip injections of 150 mg/kg D‐luciferin (Xenogen) 10 min before imaging. To quantitate tumor burden, bioluminescence signals were calculated from the imaging data using the Living Image software 3.2 (Xenogen) according to manufacturer's protocols.

3. Results

3.1. Anti‐proliferative potential of indolinones on HCC cell lines

Ten benzylidene‐indolinones functionalized at R′ and R″ (Table 1) were synthesized. The anti‐proliferative potential of these compounds was initially tested over a concentration range of 0.1–10 μM on two representative liver cancer cell lines HepG2 and HuH7. THLE2 was employed as a normal hepatocyte cell line to differentiate anti‐cancer activity from non‐specific cytotoxic effects. IC₅₀ values were derived from 50% inhibition of cell viability at 72 h. As shown in Table 2, 47 exhibited sub‐micromolar IC₅₀ against both HuH7 and HepG2 (0.5 μM and 0.6 μM respectively). Compounds 46 and 48 displayed similar potencies on HepG2 but were less potent on HuH7 (Table 2). These responses were either comparable or superior to sunitinib, which had IC₅₀ values of 4.7 μM and 4.5 μM in HuH7 and HepG2, respectively.

Table 2.

Indolinones synthesized. Compounds 39–48 were synthesized with substituents R′ and R″ on benzylidene‐indolinone as shown below.

Number	R′	R″	Number	R′	R″
39	3′‐OCH3	6‐F	44	3′‐OH, 4′‐OCH3	6‐Cl
40		5‐Cl	45		6‐OCH3
41		6‐Cl	46	3′‐CF3	6‐F
42		6‐OCH3	47		6‐Cl
43	3′‐OH, 4′‐OCH3	6‐F	48		6‐OCH3

Table 1.

IC50 of test compounds based on viability assay in HuH7, HepG2 and THLE2. Cell viability assay shown in Figure 1 was analyzed by Prism5 to determine respective IC50 values.

Compound no.	IC50 HuH7 (μM)	IC50 HepG2 (μM)	IC50 THLE2 (μM)
39	>10	7.6	>10
40	>10	9.2	>10
41	>10	8.5	>10
42	4.0	2.2	1.2
43	>10	1.2	>10
44	>10	1.0	>10
45	8.0	1.4	>10
46	1.1	0.4	>10
47	0.5	0.6	>10
48	2.3	0.4	3.6
Sunitinib	4.7	4.5	4.5

Subsequently, we quantified the selectivity ratio (SR) of tested compounds by normalizing the percentage viability at 10 μM with the normal liver control cell line THLE2/(HuH7 or HepG2). Most compounds gave a value >1, indicating enhanced anti‐proliferative effect in tumor versus normal cells (Table 3). Remarkably, 47 demonstrated a pronounced safety index when tested on both HuH7 (SR = 21.8) and HepG2 (SR = 65.3). Comparison with sunitinib and the clinically approved sorafenib confirmed that 47 exhibited better efficacy and reduced cytotoxicity (Figure 1A–C). Hence, our data emphasize 47 as a suitable candidate for further characterization.

Table 3.

Safety ratio of test compounds at 10 μM in HCC and normal liver cell lines. Cell viability was determined as described in Table 2. Percentage viability at 10 μM treatments was determined and safety ratios were calculated by normalizing survival in HCC cell lines to THLE2.

Compound no.	HuH7	SR(HuH7/THLE2)	HepG2	SR(HepG2/THLE2)
39	65.8 ± 1.1	1.2	40.0 ± 4.2	2.0
40	57.5 ± 6.6	1.6	46.1 ± 2.2	1.9
41	101.7 ± 6.9	0.9	46.4 ± 2.5	1.9
42	49.2 ± 4.0	0.8	20.7 ± 2.0	1.9
43	84.1 ± 7.5	1.2	28.3 ± 6.7	3.6
44	63.0 ± 4.0	1.4	15.0 ± 1.3	5.7
45	46.6 ± 1.8	1.5	19.9 ± 1.4	3.6
46	11.3 ± 1.0	6.4	4.2 ± 0.3	17.2
47	3.3 ± 0.2	21.8	1.1 ± 0.09	65.3
48	27.3 ± 0.5	1.3	14.5 ± 1.3	2.4
Sunitinib	3.4 ± 0.2	10.8	1.8 ± 0.3	20.4

Figure 1.

Anti‐proliferative effects of 47 on HCC cell lines (A) HuH7, (B) HepG2 and (C) THLE2 cells on 96‐well format were treated with 0‐10 µM (■), Sunitinib (●) or Sorafenib (▲) for 72 h. Viability was measured by Cell‐Titer Glo and results expressed as a percentage of viable cells in the untreated controls with ±coefficient of variation (n = 8). (D) The efficacy of 47 against HCC proliferation was tested on a wider panel of HCC cell lines: Hep3B (●), Hs817T (■), PLC/PRF/5 (▲) and SK‐Hep1 (▼), as determined by Cell Titer‐Glo assay.

3.2. Effects of compound 47 on a separate panel of HCC cell lines

Accordingly, 47 was screened against a larger panel of 4 additional HCC cell lines (Hep3B, Hs817T, PLC/PRF/5 and SK‐Hep1). These experiments corroborated its effectiveness with IC₅₀ in low micromolar or submicromolar concentrations in 3 of the 4 cell lines. SK‐Hep1 was the only marginal responder with an IC₅₀ over 10 μM. On the other hand, Hs817T cell line was most responsive to 47 with undetectable cell viability at 5 μM (Figure 1D).

3.3. Effects of compound 47 on cell cycle and apoptosis markers

Next, we investigated the biochemical mechanism of 47's anti‐proliferative effects. 47 invoked a dose‐dependent inhibition of both Erk and Akt phosphorylation at 24 h in HuH7 cells. In HepG2 cells, Akt but not Erk phosphorylation was reduced. The absence of an effect on Erk may be due to an activating N‐Ras mutation in HepG2 that drives constitutive Erk phosphorylation (Richards et al., 1990). There was also a corresponding reduction in the expression of the cell cycle marker cyclin‐D1 in both HuH7 and HepG2. Apoptotic markers like cleaved PARP (Figure 2) and caspase‐3 activation (Figure 3A) further confirmed the involvement of pro‐apoptotic mechanisms. Little effect on the mitochondrial apoptotic factors Bax and BCL‐xL was observed (Figure 2).

Figure 2.

Western blot analysis of 47 in HuH7 and HepG2 HuH7 or HepG2 cells were treated with increasing concentration of 47 (0, 1, 5 or 10 µM) for 24 h. Cell lysates were immunoblotted against cleaved PARP, phospho‐ERK, phospho‐AKT, PCNA, cyclin D1, Bax and Bcl‐xL, using HSP60 as loading control.

Figure 3.

Effects of 47 on apoptosis, AFP transcription and cell migration (A) HuH7 (darker bar) and HepG2 (lighter bar) treated as described in Figure 2 were harvested for caspase‐3 assay performed by the catalytic hydrolysis of fluorogenic Ac‐DEVD‐AMC. Results were expressed as a normalized RFU per µg lysate per h of incubation with the AMC substrate (n = 3). (B) HuH7 cells were treated with 47 (1, 5, 10 µM) or sunitinib (10 µM) for 24 h. Real‐time PCR was performed for AFP. The resulting average C _T values were normalized against 18S mRNA as housekeeping control and expressed as fold change in transcript level vs. untreated controls. Error bars represent standard deviation converted to fold changes (n = 3). (C) Serum‐starved HuH7 (200,000 ells) were seeded onto trans‐well inserts of 24‐well format and treated in triplicates as described in (B). Cells were allowed to migrate across 8 µm pores for 16 h using 10% FBS as chemoattractant. Migrated cells were stained with crystal violet and 3 fields were counted per well. Actual cell counts for respective treatments are presented with ± SD as error bars (n = 9).

3.4. Compound 47 inhibits AFP transcription in HuH7

To evaluate the specificity of anti‐tumor effect on HCC cells, the tumor marker AFP was measured on the transcriptional level. The advantage of AFP mRNA quantification over ELISA assay is its ability to determine if perturbation of AFP by 47 is due to direct gene regulation or the indirect consequence of altered cell proliferation. Using HuH7 (a high AFP‐producing cell line), AFP was significantly repressed (reduced) to one‐third (of baseline levels at 24 h after exposure to 47 (10 μM). In contrast, sunitinib showed no measurable effect (Figure 3B). We also performed a separate study to compare 47 directly with the clinically approved agent for HCC, Sorafenib, and found 47 to be superior in this aspect as well (Supplementary Data S1).

3.5. Compound 47 inhibits cell migration

Since metastasis is a major cause of mortality in HCC, we explored the possible effects of 47 on cell motility and migration using a modified Boyden chamber assay. Serum‐starved HuH7 cells incubated with 47 or sunitinib were primed to migrate using 10% FCS as chemoattractant. At 16 h, 10 μM of 47 reduced cell migration to 58% of control. Sunitinib, the clinically useful anti‐metastatic agent showed more potent inhibition at 28% of control (Figure 3C).

3.6. Profiling RTK targets by antibody array analysis

Subsequently, we investigated if RTKs were targeted by 47 and their possible contribution to efficacy in HCC cell lines. Profiling of receptor tyrosine kinase(s) inhibition by either 47 (1, 5, 10 μM) or sunitinib (10 μM) was determined by incubating lysates of treated cells with human phospho‐RTK array (RnD Systems). Higher levels of constitutively phosphorylated RTKs were found in serum‐starved HuH7 cells and thus they were used in preference to other HCC cell lines in this study (data not shown). Hence, this platform would allow us to identify distinct signaling changes upon compound treatment. Here, we observed detectable inhibition of RTKs including insulin receptor, IGF‐1R, Tyro3, EphA2, HER3, Met, RON and FGFR4 by 47 (Supplementary Data S2 and S3). The magnitude of these effects was determined densitometrically and found to be generally stronger than that achieved with sunitinib. Interestingly, EGFR displayed an increase in phosphorylation with treatment (Supplementary Data S2 and S3). Furthermore, we performed an independent verification of the inhibitory effects using a proprietary in vitro assay by Ambit Biosciences in which inhibition of a wide panel of 117 tyrosine kinases was examined. The results showed a high level of Tyro3 inhibition as compared to sunitinib and sorafenib (Supplementary Data S4). However, IC₅₀ was not achieved when tested up to 30 μM (data not shown).

3.7. Confirming IGF‐1R, EphA2 and Tyro3 phosphorylation by Western blot

Some of the significant RTK phosphorylation changes were confirmed independently by immunoblotting. From the results, proportional reduction in IGF‐1R and EphA2 phosphorylation was observed with increasing concentration of 47 (Figure 4A and 4B). This effect was specific to 47 as 10 μM sunitinib did not inhibit phosphorylation of either tyrosine kinase to any appreciable extent. While sunitinib demonstrated some IGF‐1R inhibition from the array analysis, immunoblot using phospho‐specific antibody revealed that it did not inhibit the phosphorylation sites responsible for kinase activation (i.e. Y980 and Y1135/36). On the other hand, Tyro3 phosphorylation was inhibited even at the lowest concentration tested (1 μM). Sunitinib displayed a similar effect against Tyro3. Separately, an increase in EGFR phosphorylation was also observed upon treatment with 47 but not sunitinib (Figure 4B). To confirm the immediacy of these effects on phosphorylation, these experiments were repeated at shorter exposure times (2 h instead of 24 h) to 47. Here, similar effects on RTK phosphorylation were observed (Supplementary Data S5).

Figure 4.

Effects of 47 on IGF‐1R, EGFR, EphA2 and Tyro3 (A) HuH7 cells were treated with vehicle, 47 (1, 5 or 10 μM) or Sunitinib (10 μM) for 24 h. Samples were immunoblotted against phospho‐IGF‐1R (Y980 and Y1135/1136) (B) Cells treated similarly were immunoprecipitated with anti‐EGFR, anti‐EphA2 and anti‐Tyro3 before immunoblotting with anti‐phosphotyrosine (4G10). Membranes were subsequently reprobed with respective antibodies as loading controls. (C) HuH7 cells were transfected with scrambled, IGF‐1R or EphA2 siRNAs (50 nmol) for 72 h before treating with 47 (10 μM) for 24 h and harvested for caspase‐3 assay as described for Figure 3A. Results were expressed as a normalized RFU per μg lysate per h of incubation with the AMC substrate (n = 3).

To further confirm that the biological effects of 47 are mediated by multi‐kinase inhibition, the loss in cell viability after treatment was compared with that caused by established IGF1R inhibitors. 47 significantly inhibited viability at much lower concentrations than either AG1024 or PPP (picropodophyllin) in HuH7 and HepG2 cell lines (Supplementary Data S6).

3.8. Increased resistance of HuH7 cells to 47 after IGF‐1R, Tyro3 and EphA2 silencing in HuH7

To determine if the RTKs inhibited by 47 contributed to the observed biological effects, the potential targets of IGF‐1R, EphA2 and Tyro3 were silenced by the siRNA approach and the effect on apoptosis was measured. Four siRNA sequences were tested for each gene and the ones with the best knock‐down efficiency (>70% reduction in protein levels) were selected for the assay. Tyro3 was not effectively silenced by all constructs tested and was hence excluded from subsequent analysis (Supplementary Data S7). After 72 h of gene silencing, the surviving cells were re‐plated and treated with 47. Using caspase‐3 activity as an apoptotic marker, we observed that the cells with IGF‐1R and EphA2 knocked down, gained resistance to 47 as compared to non‐silenced cells. Both IGF‐1R and EphA2 silenced HuH7 cells demonstrated a reduction of apoptosis by approximately 50% in each case (Figure 4C).

3.9. Compound 47 suppressed in vivo tumor growth in an orthotopic HCC model

We tested the antitumor potential of 47 in vivo using an orthotopic HCC model with the human HuH7‐Luc tumor cells. Mice were treated either with DMSO alone, with ip injection of 47 at 60 mg/kg or sorafenib at 30 mg/kg daily oral dose. Tumor growth was monitored up to 20 days. Quantitative analysis at set time points were quantified by measuring photon counts and expressed as tumor burden relative to photon counts before the first therapeutic injection. It was determined that 47 at 60 mg/kg induced significant inhibition of tumor growth compared with the DMSO‐treated controls, P < 0.0001 at day 20 (Figure 5A and B). Representative bioluminescence images are as shown in Figure 5C, indicating 47 clearly impeded tumor growth in a manner comparable, if not better, than sorafenib‐treated mice (Figure 5B and C). No significant change in body weight was observed in the 47‐treated mice (Figure 5D).

Figure 5.

Effects of 47 on orthotopic tumor growth HuH7_Luc tumor cells were used to induce orthotopically implanted HCC tumors in Balb/c nude mice as described in Materials and methods. (A) Tumor growth in Balb/c nude mice bearing orthotopically implanted HuH7‐Luc tumor treated with either the DMSO solvent, 47 at 60 mg/kg once every week via ip injection or 30 mg/kg sorafenib daily via oral gavage. Tumor burden was calculated. Point = mean; bars = SE. (B) Statistical comparison of tumor growth of control, 47‐ and sorafenib‐treated mice on day 20. (C) Representative bioluminescent images of control and treated mice on day 3 and day 20 following treatment. (D) Graphs showing the body weight of individual mouse in the control, 47‐ and sorafenib‐treated group.

4. Discussion

While a number of tyrosine kinases have been linked to HCC, few have translated into clinically useful therapy. Existing TKIs in HCC trials are primarily agents optimized for the treatment of other cancers and thus may not exhibit the best kinase inhibitory profile to counteract the signaling abnormalities that are characteristic of HCC. This limitation, coupled to the likely involvement of multi‐genic lesions in HCC, led us to explore multi‐targeted kinase inhibitor(s) for greater treatment efficacy. Hence, we employed a biological phenotypic screen using HCC cell lines to improve the chance of isolating target candidate(s) with greater relevance for liver cancer.

We screened benzylidene‐indolinones using inhibition of cell viability in HuH7 and HepG2 as preliminary indications for efficacy. The benzylidene‐indolinone ring is a versatile scaffold with multiple sites for derivatization. It also satisfies key drug‐like properties such as small molecular weight (<500) and favorable lipophilicity (c Log P < 5) for permeability and distribution (Lipinski, 2000). This scaffold has already led to approved agents like sunitinib and other candidates under evaluation like BIBF1120 and SU5416. Therefore, we reasoned that this scaffold may have a favorable recognition profile in the ATP‐binding pocket of multiple kinases, and that subtle modification of the scaffold may help uncover other analogs of clinical value.

Here, we found a subset of substituted benzylidene‐indolinones compounds 46–48 to have good growth inhibitory activities on HCC cell lines. Interestingly, these compounds share the common feature of a 3′‐trifluoromethyl substituent on ring B. In particular, 47 inhibited cell proliferation at submicromolar IC₅₀ and is superior to both sunitinib and sorafenib, in terms of suppressing the HCC marker, AFP. This efficacy could be reproduced across a range of HCC cell lines. The only weak responder was the lone mesenchyme‐like cell line, SK‐Hep1 which has a very different pathology (secondary tumor from a distant‐metastasis of colon carcinoma). Additionally, 47 exhibited a favorable safety profile as shown by significantly reduced cytotoxicity in non‐cancerous THLE2 liver cell line. This is a critical attribute for multi‐targeted kinase inhibitors because the inherently low stringency in target selectivity could inadvertently lead to more off‐target effects. Currently, hepatotoxicity is the major dose‐limiting toxicity for many TKIs and a good selectivity for the tumor cells would help to circumvent such potential adverse events (Teo et al., 2013). Furthermore, 47 demonstrated significant pro‐apoptotic and anti‐migratory activity that enhanced the breadth of biological effects required of a good anti‐cancer agent. Separate effort also led to characterization of it anti‐angiogenic effect, based on an HUVEC tube formation assay (unpublished data by personal communication).

As the preceding results supported 47 as a promising lead candidate, we retrospectively investigated the kinases inhibited by this compound, with the objective of deducing the mechanistic basis for its activity in HuH7. We are aware that such a strategy would not lead to a comprehensive identification of all kinases targets, but it could pinpoint the constitutively phosphorylated RTKs in HCC that are more likely to play pivotal roles in sustaining tumorigenic progression. Accordingly, we noted a subset of RTKs being inhibited in HuH7. The kinases with high levels of phosphorylation that were effectively inhibited by 47 included IGF‐1R, EphA2 and Tyro3. Yet, these targets were not strongly inhibited according to in vitro kinase assay, suggesting that the potent cell‐based inhibition could be an effect of an indirect inhibition, or through binding to a site apart from the ATP binding site. Nonetheless, these kinase targets would provide useful proof‐of‐concept support for our hypothesis that multi‐kinase inhibition was responsible for the pharmacological effect of 47. Interestingly, we also observed a paradoxical increase in EGFR phosphorylation. Although EGFR activation commonly correlates with increased cell proliferation through ERK signaling, this downstream effect was not detected (Figure 2). We speculate that EGFR activation may be a compensatory response to IGF‐1R inhibition as the tumor attempts to switch oncogenic dependence to an alternative pathway. The failure of ERK phosphorylation arising from EGFR activation suggests that IGF‐1R, and not EGFR, has overriding control on tumor proliferation. The inter‐dependence between EGFR and IGF‐1R is supported by previous reports of cross‐talk where hepatoma cells become resistant to the EGFR inhibitor, gefitinib through IGF‐1R signaling (Desbois‐Mouthon et al., 2006). A reciprocal inhibition of IGF‐1R also resulted in an increase in EGFR phosphorylation in BxPC3 cells which corroborated with our observations (Buck et al., 2008). Likewise, another study described that perturbation of IGF‐1R directly altered ERK phosphorylation in HuH7 cells and contributed to oncogenesis (Cheng et al., 2008), thus affirming IGF‐1R as the more likely transducer of MAPK signaling in HCC in contrast to the dominant role of EGFR signaling seen in other cancer types.

IGF‐1R is an oncogene of increasing significance in cancer research. Reports have described the role of IGF‐1R in hepatocarcinogenesis and its control over downstream cell cycling and anti‐apoptotic pathways in HCC models (Cheng et al., 2008; Hopfner et al., 2006; Scharf and Braulke, 2003). Our observation of IGF‐1R being strongly phosphorylated (besides the physiologically active insulin receptor) in HuH7 qualifies it as a key candidate for the maintenance of the HCC phenotype. Having said that, the superior effects we observed with 47 compared to potent IGF1R inhibitors (AG1024 and PPP) suggest that additional kinases may also contribute to the efficacy. Thus, the concurrent inhibition observed for constitutive EphA2 and Tyro3 phosphorylation in HuH7 by 47 strongly suggests their potential as novel oncogenes in HCC. EphA2 is an emerging target as accumulating evidence suggests its overexpression and role in malignancies such as melanoma (Walker‐Daniels et al., 2003). A recent report demonstrated that EphA2 was elevated in lymph metastasis of HCC, providing further support for its putative role as an oncogenic target (Lee et al., 2009). Tyro3 belongs to the Axl subfamily of RTKs (also known as SKY). To date, there are only limited reports on the association of Tyro3 with human cancers although its transformation properties have been experimentally demonstrated in experimental models (Hafizi and Dahlback, 2006). While the exact function of these RTKs in HCC requires further investigation, the inhibition of the cell cycling marker Cyclin D1 and the stimulation of early apoptosis (caspase‐3 activation and PARP cleavage) provide mechanistic support that multi‐kinase inhibition affects signaling changes that could lead to synergistic growth arrest and apoptotic cell death.

Insulin receptor was inhibited by 47 but to a significantly lower extent than Sunitinib, which interestingly has been reported to have anti‐diabetogenic effect (Louvet et al., 2008). This may account for the reduced cytotoxicity observed with 47 compared to Sunitinib, given the many physiological functions insulin‐dependent signaling exerts on the liver. Yet, emerging evidence has linked elevation in insulin receptor levels to various cancers and may participate in cell de‐differentiation (Frasca et al., 2008; Pollak, 2008). Of note is our observation that the insulin receptor is the most strongly phosphorylated RTK in our serum‐starved HuH7 cells. This raises the possibility that dysregulated signaling through this pathway may also contribute to HCC growth. Hence, inhibition of insulin receptor cannot be excluded as a potential anti‐cancer target for the disease.

While we have not explored the complete spectrum of kinase targets for 47, the evidence provided so far highlight the marked differences in the kinase inhibitory profiles of 47 and the classical multi‐RTK inhibitors, sunitinib and sorafenib. The inhibition of IGF‐1R and EphA2 by 47 was not seen with sunitinib. The involvement of different targets could further account for the discrepancy in AFP regulation where 47 but not sunitinib, dramatically depleted AFP transcription. This finding alluded to the possibility that the panel of kinases inhibited by 47 may be involved in the regulation of AFP, thereby conferring a selective effect on HCC, where 60–70% of the patients exhibit elevated AFP (Abelev and Eraiser, 1999). The therapeutic advantage of this effect is enhanced by a recent study reporting that AFP knock‐down promoted apoptosis in HuH7 cells, suggesting that AFP may be actively regulating HCC growth and not just playing the role of an innocuous tumor marker (Yang et al., 2008). Most importantly, we have demonstrated the in vivo efficacy of 47 in shrinking tumors using physiologically relevant orthotopic HCC rodent model. This effect is comparable to sorafenib, even though 47 was administered at a low frequency of once weekly. Therefore, opportunities are presented to further optimize the treatment response of 47 with subsequent preclinical pharmacokinetics studies. Overall, multi‐kinase inhibition makes 47 an interesting and potentially useful agent for lead optimization and characterization of the RTKs involved in the progression of HCC. The divergence of RTK targets inhibited by 47 from those affected by sunitinib could be advantageous, given recent reports that sunitinib and other VEGF‐specific therapy may promote a paradoxical tumor metastasis despite its anti‐angiogenic properties (Ebos et al., 2009; Paez‐Ribes et al., 2009).

In conclusion, we have identified a new indolinone derivative that displays significant safety and efficacy advantages over existing TKIs from in vitro models. The inhibition of major hyperphosphorylated RTKs in HCC underscore their roles in maintaining the progression of HCC and would be investigated in greater depth. Therefore, multi‐targeted kinase inhibitors acting on these RTKs may represent a new class of TKIs for use against HCC.

Financial disclosure

The authors who have taken part in this study declared that they do not have anything to disclose regarding funding or conflict of interest with respect to this manuscript.

Supporting information

The following is the supplementary data related to this article:

Supplementary data

Supplementary data 1.

Supplementary data related to this article can be found in the online version at http://dx.doi.org/10.1016/j.molonc.2014.04.008.

Ho Han Kiat, Chua Boon Tin, Wong Winnie, Lim Kah Suan, Teo Vivian, Ong Hooi-Tin, Chen Xiao, Zhang Wei, Hui Kam M., Go Mei Lin, Ullrich Axel, (2014), Benzylidene‐indolinones are effective as multi‐targeted kinase inhibitor therapeutics against hepatocellular carcinoma, Molecular Oncology, 8, doi: 10.1016/j.molonc.2014.04.008.