A Gold(III) Porphyrin Complex with Antitumor Properties Targets the Wnt/β-catenin Pathway

Breast cancer represents the most common diagnosed female malignancy and the second leading cause of women death worldwide (1). Chemotherapeutic agents are commonly used and usually given in the form of combinational chemotherapy (2). However, the effects of these agents are not universal, and a large portion of patients will develop resistance. Moreover, side effects including induction of life-threatening toxicity are commonly encountered.

Therapeutic values of gold have been recognized thousands of years ago, and its rational use in medicine began in early 1920s (3–7). Because gold(III) is isoelectronic with platinum(II) and tetracoordinate gold(III) complexes are in the same square-planar geometries as cisplatin, the potential anticancer properties of gold(III) complexes have been investigated for almost three decades (8, 9). A number of gold(III) complexes have been reported to exhibit cytotoxicities against a broad spectrum of tumor cells, and their potencies (IC₅₀ values in low micromolar range) are comparable with that of cisplatin. In contrast to general expectations, evidence suggest that the gold(III) complexes exert their antiproliferative activities through mechanisms that are substantially different from those of platinum drugs (10). Yet, the molecular mechanisms and targets of gold(III)-based antitumor metallodrugs remain largely uncharacterized.

Previously we synthesized a series of gold(III) meso-tetraarylporphyrin complexes characterized by enhanced stability in aqueous solutions and under physiological conditions (11). Among them, gold-1a [5,10,15,20-tetraphenylporphyrinato gold(III) chloride] showed promising antiproliferative activities against human cancer cells, including those derived from neuroblastoma, colon, nasopharyngeal, and hepatocellular carcinomas (11–15). In the present study, a novel gold(III) porphyrin analogue [5-hydroxyphenyl-10,15,20-triphenylporphyrinato gold(III) chloride (gold-2a)] was prepared by modifying one of the peripheral phenyl groups of the gold-1a with a hydroxyl substitution to improve its aqueous solubility (Supplementary Fig. S1). The efficacy of gold-2a in suppressing the growth of a panel of human breast cancer cells was evaluated in vitro, and its anticancer activity was investigated in nude mice. Our results suggest that gold-2a is a promising drug lead for anti–breast cancer treatment and that it can selectively inhibit Wnt/β-catenin signaling through modulating histone deacetylase (HDAC) activities.

The synthesis of the gold(III) porphyrin complex (gold-2a) was conducted under a nitrogen atmosphere using the standard Schlenk technique (11). ¹H nuclear magnetic resonance (NMR) spectrum was recorded on a DPX-400 Bruker FT-NMR spectrometer with chemical shift (in ppm) relative to tetramethylsilane. Absorption spectrum was recorded on a Perkin-Elmer Lambda 900 UV-vis spectrophotometer. Mass spectrum (FAB) was recorded on a Finnigan MAT95 mass spectrometer using 3-nitrobenzyl alcohol (NBA) as the matrix. Elemental analyses were performed at the Institute of Chemistry at Chinese Academy of Sciences, Beijing. The gold-2a was obtained at a yield of 77%. ¹H NMR (CDCl₃): δ = 9.44 (d, J = 5.2 Hz, 2H), 9.30−9.25 (m, 6H), 8.23 (d, J = 6.3 Hz, 6H), 7.93-7.79 (m, 11H), 7.32 (d, J = 8.4 Hz, 2H). UV-vis (DMSO) λ_max/nm (log ϵ): 414 (5.35), 526 (4.30). FAB-MS: m/z 826 [M⁺]; elemental analysis calculated (%) for C₄₄H₂₈N₄OClAu: C, 61.37; H, 3.28; N, 6.51. Found: C, 61.54; H, 3.17; N, 6.46. For cell culture and other in vitro experiments, gold-2a and cisplatin were dissolved in DMSO, whereas for in vivo animal administration, these complexes were dissolved in PET (a mixture of 60% polyethylene glycol, 30% ethanol, and 10% Tween-80) and further diluted with PBS at a 1:9 ratio.

Cell proliferation was evaluated by crystal violet staining and [³H]thymidine incorporation methods as described previously (16, 17). The human breast carcinoma BT474, MCF-7, MDA-MB-231, SKBR3, T47D cell lines were obtained from American Type Culture Collection on December 2004. The identities of these cells were confirmed by STR-profiling using the Cell ID System from Promega (last tested on 16/02/09).

MDA-MB-231 cells (5 × 10⁶) were implanted into the third right thoracic mammary fat pad of female nude mice (6 wk old) as described before (17). Six days after the implantation, the mice were divided into groups consisting of animals bearing almost the same size of tumor. Intraductal administration of the drugs was performed at the 7th and 11th days after initial implantation (day 0). I.p. drug administration was performed twice weekly as indicated. Tumor development was monitored every 3 to 4 d using digital vernier calipers, with tumor volume calculated using the formula [sagittal dimension (mm) × cross-dimension (mm)]² / 2 and expressed in mm³. All animal experimental protocols were approved by the animal ethics committee at the University of Hong Kong.

Nuclear activities of endogenous β-catenin were analyzed by TOPflash/FOPflash reporter system (17).

Quantitative reverse transcription–PCR (RT-PCR) was performed as described (17). The primers were listed in Supplementary Table S1. The histone acetylation at the promoter regions of different genes was compared by chromatin immunoprecipitation (ChIP)–PCR and ChIP-QPCR. Briefly, 8 × 10⁶ tumor cells were cross-linked with 1% formaldehyde and sheared by sonication to obtain chromatin with an average DNA length of 500 to 1,000 bp. After a 3-h preclear with protein G–Sepharose beads (blocked with 0.2 mg/mL salmon sperm DNA and 0.5 mg/mL bovine serum albumin), 100 μg chromatin were incubated with 2 μg of antibodies at 4°C overnight. Immunocomplexes were collected by protein G–Sepharose (blocked as above), washed, and eluted in 50-μL elution buffer (0.1 mol/L NaHCO₃, 1% SDS). After reversed cross-linking, the chromatin fragments were purified by phenol-chloroform extraction. PCR amplifications were performed using 10% of the DNA samples, and the resulting products were quantified by densitometry (MultiAnalyst Software, Bio-Rad). The primers were listed in Supplementary Table S2. Quantitation was also performed using real-time PCR analysis.

HDAC activity was determined using the colorimetric HDAC activity assay kit (BioVision, Inc.) and the Fluor de Lys Substrate (BML-AK500, BIOMOL International) according to the manufacturer's instructions. Briefly, for in vitro HDAC activity determination, 150 μg of MDA-MB-231 nuclear extracts were incubated with different dosages of gold-2a for various periods of time. The reaction was initiated by addition of HDAC colorimetric substrate and incubation at 37°C for 1 h. The developer was then added, and the samples were incubated at 37°C for another 30 min before reading at 405 nm. Cells pretreated with drug compounds were also processed for HDAC activity measurement. In addition, the inhibitory effects of gold-2a on individual HDACs were tested using immunoprecipitated complexes.

Histones were extracted according to the procedure reported previously (18). The acid soluble histone fraction was subjected to 18% SDS-PAGE and Western blotting analysis.

Subcellular fractionation was performed as described (19). Nuclear fractions were further separated into nuclear matrix, nucleoid, and nucleolus. All samples were subjected to inductively coupled plasma–mass spectrometry (ICP-MS) analysis (7500 series, Agilent Technologies, Inc.) for determining the content of the stable gold isotope ¹⁹⁷Au (20).

The 1.9-Å X-ray structure of HDAC8 (Protein Data Bank code 1t64) was used for docking calculations using Gaussian 03. The gold-2a complex was optimized using DFT with a LanL2MB basis set (21). ICM-Pro 3.6-1 program (Molsoft) was applied for molecular docking. Energy calculations were based on the ECEPP/3 force field with a distance-dependent dielectric constant (22). The biased probability Monte Carlo minimization procedure was used for global energy optimization. All ICM dockings were performed thrice, and a minimum of the three interaction energies was used.

All experiments were performed with three to eight samples per group, and all results were derived from at least three independent experiments. Data are shown as mean values ± SD. Unless otherwise specified, comparison between groups was done using Student's t test. P < 0.05 was used to indicate a significant difference.

Five types of human breast carcinoma cells with distinct gene expression profiles and oncogenic phenotypes, including BT474, MCF-7, T47D, MDA-MB-231, and SKBR3 (23), were used for evaluating the cytotoxicity of gold-2a. Crystal violet assay showed that gold-2a inhibited growth of all five types of cells at mean IC₅₀ values of 0.49 ± 0.17, 0.08 ± 0.04, 0.04 ± 0.01, 0.007 ± 0.002, and 0.02 ± 0.01 μmol/L, respectively [0.5% fetal bovine serum (FBS) condition, 24-hour treatment; Table 1]. The presence of high concentrations of serum had no effects on the potencies of the drug. In contrast, the IC₅₀ values of cisplatin were ∼100 to 3,000 times higher than gold-2a. Similar results were observed when the drug exposure time was extended to 48 hours. It should be noted that the IC₅₀ values of gold-2a were one to two logs lower in MDA-MB-231 cells than the other four types of cells under all conditions. On the other hand, the potency of gold-2a toward noncancerous fibroblast cell was ∼10-fold to 600-fold lower than those of mammary cancer cells with an IC₅₀ of 4.17 ± 1.67 μmol/L. The antiproliferative activity of another gold(III) porphyrin complex {[Au^III(TPP)]Cl, gold-1a} in MDA-MB-231 cells was also tested. Nonlinear regression analysis of the growth inhibition curves revealed that gold-1a was ∼450-fold less effective than gold-2a (IC₅₀, 1.351 ± 0.032 μmol/L versus 0.003 ± 0.008 μmol/L). Apoptosis of MDA-MB-231 cells were evaluated by measuring DNA fragmentation and terminal deoxynucleotidyl transferase-mediated dUTP nick end labeling analysis. The results showed that gold-2a significantly increased DNA fragmentation and the number of apoptotic cells in a dose- and time-dependent manner (Supplementary Fig. S2).

To evaluate the in vivo antitumor effects of gold-2a, MDA-MB-231 cells were implanted into athymic nude mice and different drug dosages were tested for the treatment. Biweekly i.p. administration of gold-2a (1.5, 3.0, and 6.0 mg/kg) for up to 4 weeks dose-dependently attenuated the tumor growth (Supplementary Fig. S3). However, no complete tumor suppression could be achieved despite the fact that the animals tolerated the treatment well. Instead, the administration of two bolus of the gold-2a (15 mg/kg) by intraductal injection into tumor xenograft resulted in a complete tumor remission in 50% of the animals at 2 weeks after initial implantation (Fig. 1A). Most of the animals remained in a tumor-free status until day 25, at which recurrence of tumor was observed. Comparing to gold-2a, intraductal delivery of the same dosage of cisplatin attenuated the rates of tumor growth to a much less degree. Of note is that ∼40% of animals died in the cisplatin treatment group after two injections, whereas those of the PET control and gold-2a treatment groups remained alive during the experimental periods (Fig. 1B). The body weights of both gold-2a– and cisplatin-treated mice were slightly lower than the PET control group (Fig. 1C). The average tumor weights in gold-2a and cisplatin groups were 0.17 ± 0.02 and 0.42 ± 0.15 g, respectively, which were decreased by 73% and 34% when compared with control group (Fig. 1D).

Aberrant activation of the Wnt/β-catenin signaling and intracellular accumulation of β-catenin protein have been observed in a large portion of human breast tumors (24, 25). Gold-2a decreased the protein levels of β-catenin in MDA-MB-231 cells as early as 4 hours after treatment (Supplementary Fig. S4A). Nuclear transcriptional activities of β-catenin were also dramatically reduced by gold-2a (Supplementary Fig. S4B). On the contrary, cisplatin had no influence on both the protein levels and nuclear activities of β-catenin. In the absence of a Wnt signal, β-catenin is phosphorylated by glycogen synthase kinase-3β, which facilitates the subsequent ubiquintination and proteasome degradation. However, gold-2a treatment did not alter the relative phosphorylation or ubiquintination levels of β-catenin and had no obvious effects on proteasome activities (Supplementary Fig. S5). In the mean time, decreased phosphorylations of Akt and GSK-3β were observed from 6 hours onward after gold-2a treatment, later than the effects on β-catenin protein levels (Supplementary Fig. S4A). To address the cytotoxic mechanisms of gold-2a, ICP-MS was performed for monitoring the intracellular localization of this complex. The results showed that gold-2a rapidly (within 30 min) entered the nuclei of MDA-MB-231 cells and was enriched in the nucleoid fractions (Supplementary Fig. S4C).

The drop in intracellular levels of β-catenin can be triggered by a loss of Wnt signal–induced stabilization (26). In fact, quantitative PCR analysis revealed that gold-2a treatment time- and dose-dependently altered the mRNA expressions of several Wnt signaling molecules, including WNT1, WNT5B, WIF1, WISP1, and CTNNB1 (Fig. 2). For instance, gold-2a treatment largely blocked the expression of WNT1 but profoundly augmented the mRNA levels of WIF1. Significant upregulation of WIF1 (over 80-fold) was observed at as early as 2 hours, and the stimulatory effects continued during the 24-hour course of treatment. The mRNA levels of WISP1 were rapidly decreased to an undetectable level after gold-2a treatment. The significant effects of gold-2a on CTNNB1 and WNT5B expression were observed from 6 hours of treatment onward. Similar trends of changes were also shown for the protein expressions of WNT1, WNT5B, WIF1, and WISP1 (Fig. 2). These results indicated that gold-2a might elicit its antiproliferative effects through regulating the gene transcriptions of Wnt/β-catenin signaling molecules.

The above results showed that gold-2a could selectively enhance the gene expression of WIF1 and noncanonical WNT5B but inhibit those of WNT1, CTNNB1, and WISP1. Epigenetic regulations, such as DNA methylation and histone acetylation, represent important mechanisms for the aberrant activation of Wnt signaling during cancer development. The results in Fig. 3 showed that gold-2a exhibited potent inhibitory effects on the enzyme activities of HDAC, the dynamic transcriptional regulator for deacetylating chromatin histones. Treatment with gold-2a resulted in a rapid decrease of the HDAC activities in MDA-MB-231 cells (Fig. 3A). The inhibitory effects could also be observed by coincubation of gold-2a with the nuclear extracts derived from untreated MDA-MB-231 cells (Fig. 3B). The HDAC inhibition was proportional to the incubation time, and the potency of gold-2a was comparable with that of trichostatin A (TSA), a specific inhibitor for multiple HDACs (27, 28). To test whether gold-2a could act as a preferential inhibitor, individual HDAC (HDAC1–HDAC9) was immunoprecipitated from the MDA-MB-231 cells and incubated with gold-2a, which was found to be able to inhibit the activity of all class I HDACs, including HDAC1, HDAC2, HDAC3, and HDAC8 (Fig. 3C).

To further confirm these unexpected observations, ChIP-PCR was performed to quantify the acetylated histone H4 levels at the promoter regions of the above genes. As the results shown in Fig. 4, treatment with gold-2a enhanced the binding of acetylated histone H4 to WIF1 and WNT5B promoters. In contrast, a decrease in acetylated histone H4 was observed at WNT1 promoter after gold-2a treatment. TSA had no significant effects on these three genes but increased acetylated histone H4 binding to the promoter of estrogen receptor α (ESR1). The acetylated histone H4 binding to the promoter of β-catenin gene (CTNNB1) and WISP1 was not significantly altered by gold-2a or TSA treatment. Western blotting analysis showed significant induction of histone H4 acetylation by TSA, which occurred as early as 6 hours after treatment (Fig. 4). Gold-2a was less effective on global histone H4 acetylation. Similar levels of acetylated histone H4 were only observed at 24 hours after treatment.

Among the four members of class I HDACs, HDAC2 could not be detected at the promoters of all five Wnt pathway genes, despite that it was present at the promoter of ESR1. The abundance of HDAC1, HDAC3, and HDAC8 was highly variable at the promoter regions of these genes (Fig. 5). HDAC1 was more abundantly associated with WNT5B than other promoters. HDAC3 was not associated with WNT5B promoter. The amounts of HDAC3 and HDAC8 at WIF1 promoter were much higher than the other four genes. The highest and lowest association could differ by as much as 10,000 times. With this information, the significant fold increases that were below 10 may not be biologically important. Gold-2a, to a greater extent, elicited inhibitory effects on the associations of different HDACs to the promoters of WIF1, WNT1, WNT5B, and CTNNB1 (Fig. 5). In samples treated with 1 μmol/L gold-2a, HDAC1 at the promoters of WIF1 and WNT5B was dramatically decreased by 54- and 164-fold, respectively; HDAC3 at the promoters of WIF1, WNT1, and CTNNB1 was downregulated by 193-, 23-, and 162-fold, respectively; the binding of HDAC8 to WIF1 promoter was also found to be downregulated by ∼300-fold.

To understand how gold-2a interacts with class I HDAC, molecular docking studies were performed using flexible-ligand docking module of ICM-Pro 3.6-1 based on the known X-ray crystal structures of HDAC8 (Supplementary Fig. S6). As a reference, TSA showed an interaction energy of −10.28 kcal/mol (Supplementary Fig. S6A and C). The root mean square deviation was within 1.0 Å when compared with the X-ray crystal structure of HDAC8-TSA (PDB code: 1T64). Comparative analysis of the low-energy conformations suggested a strong interaction between gold-2a and the binding pocket at the surface of HDAC8 (binding energy of −9.67 kcal/mol; Supplementary Fig. S6B and D). Unlike the binding modes of TSA, the side chains of gold-2a are not buried inside the 11-Å channel. The porphyrin rings are filled into the hydrophobic pocket and in close contact (within 4 Å) with amino acid residues Tyr100, Phe152, Gly151, His180, Met274, and Pro273. The OH group of gold-2a points toward the 14-Å internal cavity of HDAC8, which blocks the exit of the 14-Å internal cavity of HDAC8.

The presence of nuclear β-catenin in a broad spectrum of human cancers makes Wnt/β-catenin pathway an attractive target for cancer treatment. Nevertheless, the development of such therapeutics is still in its infancy, with no options currently undergoing late-stage clinical trials (29). In breast cancer, the Wnt/β-catenin pathway is mostly deregulated by loss of negative Wnt signaling regulators through epigenetic silencing (30). For example, the Wnt antagonists, such as WIF1 and secreted frizzled-related protein 1 (sFRP1), are inactivated due to promoter hypermethylation (31, 32). The present study showed that the drug complex gold-2a decreased the protein and mRNA levels of β-catenin and inhibited its nuclear transcriptional activities. Whereas the expression of sFRP1 was extremely low and could not be upregulated by gold-2a, the mRNA levels of WIF1 were rapidly and dramatically augmented upon drug treatment. On the other hand, the expressions of WNT1, CTNNB1, and WISP1 were abolished by gold-2a. These effects are attributed, at least partly, to the inhibitory actions of gold-2a on class I HDAC activities, which differentially regulate the histone acetylation at the gene promoters of these Wnt signaling molecules. The effects of gold-2a on gene transcription are consistent with its rapid entry into the cell nuclei.

Histone acetylation and deacetylation play fundamental roles in modulating chromatin topology and gene transcription (33). Deacetylation is a process mediated by HDACs, which remove acetyl group from the histone tails, causing the histones to wrap more tightly around the DNA and interfere with gene transcriptions. HDACs are often overexpressed in tumors compared with surrounding normal tissue and correlated with poor prognosis and survival rate (34). HDAC inhibitors (HDACi) are a promising new class of anticancer agents that can inhibit proliferation and induce differentiation and/or apoptosis of tumor cells, with little toxicity to normal cells (35, 36). Several HDACis are currently in phases I and II clinical trials on hematologic and solid malignancies (37–39). Results in the present study suggested gold-2a to be a potential HDACi. It dose- and time-dependently inhibited the HDAC activities in both cell cultures and isolated nuclear extracts and also regulated the global histone acetylation levels, despite of a lower efficiency than TSA. Although the overall result of deacetylation is a global (nonspecific) reduction in gene expression, it has been found that HDACi can selectively modulate a certain set of genes (36, 40). Here, the results showed that gold-2a regulated promoter histone acetylation in a different manner from that of TSA. Gold-2a was more effective on Wnt/β-catenin signaling genes but had less effect on ESR1. Although the detailed mechanisms underlying the HDACi effects of gold-2a were not fully addressed in the present study, experiments using individual immunoprecipitated HDAC complexes revealed that this drug complex might act as a preferential inhibitor for class I HDACs. The proposed mechanism of gold-2a as a class I HDACi was further supported by molecular modeling of the binding interactions with HDAC8 (with favorable calculated binding energy of −9.67 kcal/mol). Although the three-dimensional structures of all HDACs have not been fully resolved, the available evidence suggest that the hydrophobic pockets on their surface near the exit of the 14-Å channel show the most flexibility comparing with other regions (41). Gold-2a occupies almost all pockets of HDAC8 and blocks the entry of the 11-Å channel and the exit of the 14-Å internal cavity. On the other hand, TSA interacts with only one of these pockets. The bulk structure of gold-2a may allow stronger but more selective interactions with certain types of HDACs. Our unpublished data suggested that gold-2a may share similar mechanisms with cyclic tetrapeptide, a distinct class of HDACi (42).

Comparing with cisplatin, which had no effect on both protein stabilities and nuclear activities of β-catenin, gold-2a showed more potent cytotoxic effects in cell culture and better anticancer activities in nude mice with less systematic toxicity. The antiproliferative activities of gold-2a in MDA-MB-231 cells were comparable with other HDACi, such as SAHA (43) and TSA (44). Previous studies have suggested that cisplatin and gold (III) porphyrin complexes elicit their anticancer effects through distinct mechanisms at the DNA level and the protein level, respectively (45–48). Among the five types of human breast cancer cells, MDA-MB-231 possesses the most active autocrine canonical Wnt signaling, followed by SKBR3, T47D, MCF-7, and BT474 (16, 49), indicating that gold-2a may favor on targeting cancers that are more addicted to Wnt/β-catenin signaling. We have previously reported that gold(III) porphyrin 1a (gold-1a) can induce apoptosis through both caspase-dependent and caspase-independent mitochondrial pathways (14). The results from the present study suggest that the two compounds, although possessing similar chemical structures, might not share common cytotoxic mechanisms at the molecular level. The difference between gold-1a and gold-2a is the replacement of a benzene group by a phenol group. Subha and colleagues have reported that the presence of hydroxyl groups on test inhibitor compounds can favor the fitting into the active site of HDAC (50). It remains to be addressed whether the presence of a hydroxyl group in gold-2a assists this compound in targeting HDACs and whether the presence of more hydroxyl groups at the side arms of these gold(III) porphyrins contributes to higher toxicity or more potent inhibition toward HDACs. Further investigations are currently being carried out in our laboratory for elucidating the detailed mechanisms contributing to the selective inhibitory activities of gold-2a on HDACs.

No potential conflicts of interest were disclosed.

Grant Support: Grants from Seeding Funds for Basic Research of the University of Hong Kong (Y. Wang), Hong Kong Research Grant Council grants HKU 777908M (Y. Wang) and HKU 779707M (A. Xu), and the Area of Excellent Scheme AoE/P-10-01 established under University Grants Committee HKSAR (C. Che).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.