Histone deacetylases (HDACs) are validated targets for anticancer therapy as attested by the approval of suberoylanilide hydroxamic acid (SAHA) and romidepsin (FK228) for treating cutaneous T cell lymphoma. We recently described the bioassay-guided isolation, structure determination, synthesis, and target identification of largazole, a marine-derived antiproliferative natural product that is a prodrug that releases a potent HDAC inhibitor, largazole thiol. Here, we characterize the anticancer activity of largazole by using in vitro and in vivo cancer models. Screening against the National Cancer Institute's 60 cell lines revealed that largazole is particularly active against several colon cancer cell types. Consequently, we tested largazole, along with several synthetic analogs, for HDAC inhibition in human HCT116 colon cancer cells. Enzyme inhibition strongly correlated with the growth inhibitory effects, and differential activity of largazole analogs was rationalized by molecular docking to an HDAC1 homology model. Comparative genomewide transcript profiling revealed a close overlap of genes that are regulated by largazole, FK228, and SAHA. Several of these genes can be related to largazole's ability to induce cell cycle arrest and apoptosis. Stability studies suggested reasonable bioavailability of the active species, largazole thiol. We established that largazole inhibits HDACs in tumor tissue in vivo by using a human HCT116 xenograft mouse model. Largazole strongly stimulated histone hyperacetylation in the tumor, showed efficacy in inhibiting tumor growth, and induced apoptosis in the tumor. This effect probably is mediated by the modulation of levels of cell cycle regulators, antagonism of the AKT pathway through insulin receptor substrate 1 down-regulation, and reduction of epidermal growth factor receptor levels.
Oceans cover most of the Earth's surface and contain the greatest biodiversity in the world, unmatched by their terrestrial counterpart. Terrestrial organisms such as plants, fungi, and soil bacteria have provided mankind at least with the inspiration for approximately half of the currently marketed drugs (Newman and Cragg, 2007). The marine environment, however, is still a largely untapped resource for biomedicine. The unique marine biodiversity surely harbors unprecedented chemical structures awaiting discovery. All of these natural products have evolved to effectively modulate their respective targets and thus are biologically validated in a complex bioassay: life and survival in an intricate ecosystem. Humans are only beginning to clinically exploit some chemical secrets found in marine organisms; the first marine anticancer drug (ET-743, trabectedin/Yondelis) was recently approved in Europe (Casali et al., 2010). Salinosporamide A is a recent example of a promising marine natural product in cancer clinical trials (Fenical et al., 2009; Mayer et al., 2010).
We recently described the isolation and structure determination of largazole from a marine cyanobacterium, an antiproliferative natural product with selectivity for cancer cells over nontransformed cells (Taori et al., 2008). We devised an efficient total synthesis that can provide gram-scale quantities of largazole for rigorous biological evaluation (Ying et al., 2008b). We already established the mode of action of largazole. Largazole is a prodrug and liberates largazole thiol that can chelate Zn2+ in the active site of class I HDACs (Fig. 1) (Ying et al., 2008b). Subsequently, we have initiated structure–activity relationship (SAR) studies (Ying et al., 2008a).
HDACs regulate gene transcription by increasing the charge status of histone lysine residues through deacetylation of ε-amino groups, increasing the compactness of the chromatin complex. Conversely, inhibiting HDACs increases the accessibility of DNA to the transcriptional machinery, thereby modulating gene expression. HDACs can be grouped into Zn2+-dependent (class I, II, and IV) and NAD+-dependent (class III) isoforms (Dokmanovic et al., 2007). Class I HDAC isoforms such as HDAC1 and HDAC3 are overexpressed in various cancers, including colon, and linked to cellular proliferation (Wilson et al., 2006; Ishihama et al., 2007; Senese et al., 2007; Spurling et al., 2008). Despite seemingly nonspecific global transcriptional effects, inhibition of Zn2+-dependent HDACs has been shown to produce significant anticancer effects in vitro and in vivo. SAHA, a broad-spectrum HDAC inhibitor, and more recently romidepsin (FK228) have already been approved for treating cutaneous T cell lymphoma (Fig. 1) (Marks and Breslow, 2007; Lansigan and Foss, 2010). FK228 has superior potency and selectivity for class I isoforms and requires metabolic activation by disulfide reduction (Fig. 1). Various other HDAC inhibitors are undergoing clinical trials (Senese et al., 2007, Paris et al., 2008). Thus, we have already addressed challenges in drug development associated with marine natural products: supply problem, target identification, and target validation. Largazole has attracted tremendous interest from the synthetic and medicinal chemistry communities that rely on natural products groups to discover new promising leads, confirming our findings but also attesting to the potential of largazole and marine natural products in general (Newkirk et al., 2009; Zeng et al., 2010 and references cited therein). Here, we characterize the anticancer and HDAC inhibitory activity of largazole in cellular and in vivo cancer assay model systems.
Materials and Methods
Chemicals and Chemical Synthesis
Largazole was synthesized as described previously (Ying et al., 2008b). Largazole analogs were made by adaptation of this method, which we used previously to synthesize analogs (Ying et al., 2008a). FK228 was synthesized as described previously (Wen et al., 2008). SAHA was purchased from Selleck (Shanghai, China).
Synthesis of N-Methylated Analogs of Largazole and Synthesis of Phe, His, Asp, and Tyr Analogs of Largazole.
The synthesis of the N-methylated analogs of largazole and synthesis of Phe, His, Asp, and Tyr analogs of largazole can be found in Supplemental Schemes S1 and S2.
In Vitro Assays
Recombinant HDAC1 Enzymatic Assays.
The assays were carried out by Reaction Biology Corp. (Malvern, PA) as described previously (Ying et al., 2008a). Peptide substrate p53 residues 379 to 382 (RHKKAc), conjugated with 7-acetoxy-4-methylcoumarin, was used as the fluorogenic substrate at 50 μM assay concentration. In brief, compounds dissolved in dimethyl sulfoxide together with HDAC1 were incubated at 30°C for 2 h in the reaction buffer, which contained 25 mM Tris-HCl, pH 8.0, 137 mM NaCl, 2.7 mM KCl, 1 mM MgCl2, and 0.25 mg/ml bovine serum albumin, before the addition of developer reagent. The free acetoxy-4-methylcoumarin was detected with excitation of 360 nm and emission 460 nm in kinetic mode for 90 min. The initial velocity of an enzyme reaction was normalized and plotted with Prism (GraphPad Software, Inc., San Diego, CA) to derive the IC50 values.
Human colorectal carcinoma cells (HCT116, HT29, and HCT15) were purchased from the American Type Culture Collection (Manassas, VA) and cultured in Dulbecco's modified Eagle medium (Invitrogen, Carlsbad, CA) supplemented with 10% fetal bovine serum (HyClone Laboratories, Logan, UT) and 1% penicillin, streptomycin, and fungizone at 37°C humidified air and 5% CO2.
Cell Cycle Analysis.
HCT116, HT29, and HCT15 cells were incubated with largazole at various concentrations for 24 h. Cells were pelleted by centrifugation and fixed in ice-cold 70% ethanol. DNA was stained with 10 μg/ml propidium iodide (Invitrogen) in a reaction solution containing 100 μg/ml RNase A (Sigma-Aldrich, St. Louis, MO). Fluorescence emitted from the propidium iodide–DNA complex was quantified by using FACScan (BD Biosciences, San Jose, CA).
Cell Viability Assay.
HCT116, HT29, and HCT15 cells (1 × 104 cells per well) were seeded in 96-well clear bottom plates, and 24 h later the cells were treated with various concentrations of largazole, largazole analogs (100 pM–10 μM), or solvent control. After 48 h of incubation, cell viability was detected by using 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide according to the manufacturer's instructions (Promega, Madison, WI).
Caspase 3/7 Assays.
HCT116 cells were seeded in 96-well white assay plates (1 × 104/well), and 24 h later they were treated with largazole at various concentrations. After 48 h of treatment, the Caspase-Glo 3/7 assay (Promega) was used to measure caspase 3/7 activity. The reagent was prepared immediately before use. The lysis buffer and luciferase substrate were equilibrated to room temperature and mixed together. The assay plate was equilibrated to room temperature (∼10 min). The same volume of Caspase-Glo 3/7 reagent as culture medium was added to each well, and the plate was mixed on a plate shaker for ∼1 min and incubated at room temperature for 30 min. The luminescence was read with a luminescence plate reader (SpectraMax M5; Molecular Devices, Sunnyvale, CA).
Reverse Transcription and Real-Time Quantitative Polymerase Chain Reaction.
A total of 2 × 105 HCT116 cells (per well) were seeded in six-well dishes. Twenty four hours later, the cells were treated with largazole (20 nM), SAHA (2 μM), FK228 (20 nM), or solvent control and incubated for various time points (1–20 h). RNA was isolated with a RNeasy mini kit (QIAGEN, Valencia, CA), and cDNA was synthesized from 2 μg of total RNA by using SuperScript II Reverse Transcriptase (Invitrogen) and oligo(dT)12–18 primer (Invitrogen). Real-time quantitative polymerase chain reaction was performed in a total reaction volume of 25 μl, which included 12.5 μl of TaqMan 2× universal master mix (Applied Biosystems, Foster City, CA), 1.25 μl of 20× TaqMan gene expression assay mix, 2 μl of cDNA, and 9.25 μl of sterile water in a 96-well plate (Applied Biosystems) by using the ABI 7300 sequence detection system (Applied Biosystems). The thermocycler program consisted of 2 min at 50°C, 10 min at 95°C, and 40 cycles of 95°C for 15 s and 60°C for 1 min. Each assay was carried out in triplicate. For normalization, glyceraldehyde-3-phosphate dehydrogenase expression was used as an internal control.
RNA samples were subsequently used for GeneChip analysis. RNA concentration was further determined on a NanoDrop Spectrophotometer (NanoDrop Technologies, Inc., Wilmington, DE), and sample quality was assessed with an Agilent 2100 Bioanalyzer (Agilent Technologies, Santa Clara, CA). A GeneChip 3′ IVT Express kit (Affymetrix, Inc., Santa Clara, CA) was used for duplicate sample preparation, and the reactions were carried out according to the protocol detailed by Affymetrix. In brief, cDNA was synthesized from 250 ng of total RNA and template for in vitro transcription during which a biotin-modified nucleotide was incorporated. The biotin-labeled RNA was then purified and fragmented. Samples were hybridized with rotation at 45°C for 16 h to the Affymetrix GeneChip Human Genome U133 plus 2.0 arrays. The arrays were washed and stained with the reagents supplied in the GeneChip Hybridization Wash and Stain kit (Affymetrix, Inc.) on an Affymetrix Fluidics Station 450 and scanned with a GeneChip 7G Scanner (Affymetrix, Inc). Statistical tests were performed with Bioconductor statistical software (http://www.bioconductor.org/) and R program (R: A language and environment for statistical computing) (Gentleman et al., 2004). Raw data were normalized by the Robust Multichip Analysis approach (Bolstad et al., 2003). Then the probe set's detection call was estimated by using the Wilcoxon signed rank-based algorithm. Probe sets that are absent in all of the study samples were removed from further analyses. A linear modeling approach and the empirical Bayes statistics as implemented in the limma package (Smyth, 2004) in the R software were used for differential expression analysis. The p values obtained were controlled for multiple testing (false discovery rate) using the Benjamini-Hochberg method (Benjamini and Hochberg, 1995). Differentially expressed genes were then ranked by the p values, and genes with p value less than 0.01 and 2-fold change were considered as differentially expressed genes at a statistically significant level. Hierarchical clustering of the data were computed on log-transformed and normalized data by using complete linkage and Pearson correlation distances. Computations and visualization were done with R packages.
HCT116 cells (2 × 105 per well), HT29 cells (4 × 105 per well), and HCT15 cells (3 × 105 per well) were seeded in 60-mm dishes. The next day, largazole, SAHA, or FK228 were added to final concentrations from 0 to 10 μM and incubated for 8 or 24 h. Whole-cell lysates were collected by using PhosphoSafe buffer (Novagen, Madison, WI). The protein concentration was measured with the BCA Protein Assay kit (Pierce Chemical, Rockford, IL). Lysates containing equal amounts of protein were separated by SDS-polyacrylamide gel electrophoresis (4–12%), transferred to polyvinylidene difluoride membranes, probed with antibodies, and detected with the SuperSignal West Femto Maximum Sensitivity Substrate (Pierce Chemical). Antiacetyl histone H3 antibody was obtained from Millipore Corporation (Billerica, MA). Antiacetyl-α-tubulin, α-tubulin, β-actin, cyclin D1, EGFR, MET, HER-2, p21, and secondary (anti-rabbit, anti-mouse) antibodies were from Cell Signaling Technology (Danvers, MA). Anti-BCL2L11 was from Abcam Inc. (Cambridge, MA). Anti-cyclin-dependent kinase 6 (CDK6) and insulin receptor substrate 1 (IRS-1) were from Santa Cruz Biotechnology, Inc. (Santa Cruz, CA).
Materials and General Procedures.
HPLC-MS was done on a 3200 QTRAP (Applied Biosystems) equipped with a Shimadzu (Kyoto, Japan) UFLC System. Mouse serum and harmine were purchased from Sigma-Aldrich. Pooled CD1 mouse liver (female) microsomes were purchased from XenoTech, LCC (Lenexa, KS) with protein concentrations of 0.5 mg/ml. HCT116 cell lysates were prepared with PhosphoSafe lysis buffer (Novagen). Protein concentration was determined by using the BCA Assay.
Sample Preparation for Stability Studies.
Stock solutions of largazole and thiol were prepared by dissolving the compounds in ethanol to give a 1 mg/ml solution. Aliquots of this stock solution were then obtained to afford a 40 μg/ml solution in methanol. Serial dilution of the 40 μg/ml solution in methanol gave standard solutions with concentrations of 25, 12.5, 2.5, 1.25, 0.25, 0.125, 0.025, and 0.0125 μg/ml. A 1 mg/ml stock solution of the internal standard harmine was prepared in ethanol, which subsequently was used to prepare 100 μg/ml solution with ethanol. An aliquot of the 100 μg/ml harmine solution was diluted to 100 ng/ml with ethyl acetate to serve as the working internal standard solution.
In vitro plasma stability of largazole was done by using a modification of a published method (Chen et al., 2008). Ten microliters of largazole (25 μg/ml) was added to 100 μl of mouse serum, and the solution was vortex-mixed for 15 s and incubated for 0.25 min to 24 h (11 time points). At the end of each incubation period, 400 μl of ethyl acetate was added to each tube, followed by 200 μl of harmine to quench the reaction and extract largazole and its metabolites. Samples were further incubated in a thermomixer at 27°C (750 rpm, 5 min) and later centrifuged for 5 min at 1643g. The ethyl acetate layer was collected and evaporated to dryness under nitrogen. Samples were reconstituted in 50 μl of methanol. A volume of 10 μl of the reconstituted solution was injected into the HPLC-MS system.
The stability of largazole in the presence of mouse microsomes was determined by using an adaptation of a published procedure (Ackley et al., 2004). In brief, microsomes were added to prewarmed phosphate buffer (100 mM, pH 7.4) at 37°C. Largazole (3 μl) was added to the microsomal preparation followed by NADPH cofactor solution (1.3 mM NADP, 3.3 mM glucose 6-phosphate, 0.4 U/ml glucose-6-phosphate dehydrogenase, 3.3 mM MgCl2). The reaction was allowed to proceed for 5, 15, and 30 min at 37°C (thermomixer, 1050 rpm). The reaction was quenched by addition of ethyl acetate and subsequently spiked with harmine. The zero time point was defined by denaturing the microsomes with ethyl acetate before the addition of largazole. Incubation of largazole with microsomes alone and denatured microsomes (heated at 95°C, 5 min) with NADPH cofactor were also performed following the same procedure to determine NADPH-dependent metabolism. The maximum amount of the thiol adduct formed from largazole was defined as the amount of the metabolite formed after 5 min in the presence of microsomes alone. The final concentration of the incubation mixture contained 0.5 mg/ml protein concentration and 1 μM largazole.
Aliquots of HCT116 cell lysates were diluted with 25 mM Tris-HCl buffer, pH 8.0, to give a final reaction volume of 100 μl and protein concentration of 0.7 mg/ml. Cell lysate solutions were incubated with 10 μl of largazole (25 μg/ml) for 0.25 min to 24 h (nine time points). Remaining largazole and its metabolites were extracted from the reaction solution at the end of the incubation periods with ethyl acetate using the same procedure as described for the plasma stability assay.
The stability of largazole in aqueous solution was determined in 100 mM phosphate buffer, pH 4.0, 100 mM phosphate buffer, pH 7.0, 100 mM phosphate buffer, pH 8.0, Dulbecco's modified Eagle medium containing 10% fetal bovine serum, and 1% penicillin, streptomycin, and fungizone. Portions of each solution (100 μl) were spiked with 10 μl of largazole solution (25 μg/ml) and allowed to incubate for 0.25 min to 24 h (eight time points). The reaction was quenched at the end of each time point, and the remaining largazole was extracted after the ethyl acetate extraction procedure, as in the plasma stability study.
Analysis of largazole and its metabolites was done by using HPLC-MS [column, Onyx Monolithic C18 (3.0 × 100 mm), Phenomenex (Torrance, CA); solvent, 0.1% aqueous formic acid (solvent A) 0.1% formic acid in MeOH (solvent B); flow rate, 0.5 ml/min; detection by electrospray ionization–MS in positive ion mode (MRM scan)]. A stepwise gradient elution was used starting at 60% B and 40% A, then increasing to 100% B at 7 min and maintained at this condition for 8 min. Parameters were optimized before analysis by using direct syringe infusion. The retention times (tR, min; MRM ion pair) of the analytes and internal standard are as follows: harmine (1.5; 213 → 171), thiol adduct (2.2; 497 → 212), free thiol (4.7; 497 → 212), largazole (6.7; 623 → 497). Compound-dependent parameters used were as follows: largazole, declustering potential (DP) 30.95, entrance potential (EP) 9.16, collision energy (CE) 30.95, collision cell exit potential (CXP) 9.41, collision cell entrance potential (CEP) 58.15; thiol adduct and free thiol, DP 49.89, EP 5.32, CE 45.39, CXP 4.13, CEP 17.1; and harmine, DP 45.0, EP 9.8, CE 45.0, CXP 3.9, CEP 17.1. Source gas parameters used were as follows: curtain gas, 30.0; collision gas low, ionspray voltage 4500; temperature, 450.0; ion source gas 1 60.0; ion source gas 60.0.
Calibration curves for largazole and its metabolites in the presence of mouse serum, HCT116 cell lysates, and aqueous solutions were generated by least-square linear regression analysis of the analyte peak area and internal standard peak area ratio against the nominal concentration of the standard solutions. A linear regression analysis was performed, and the concentration of largazole, thiol, and thiol adduct at each time point was determined through interpolation for plasma, cellular, and aqueous stability experiments. The amount of remaining largazole with microsome incubation was determined from the peak area ratio of largazole at tx (5, 15, and 30 min) and t0. The amount of thiol adduct remaining during the microsomal stability assay was determined by calculating the peak area ratio of the thiol adduct at tx (5, 15, and 30 min) and t0. All calculations were done by using Analyst 1.4.2 (Applied Biosystems) Quantitate Mode.
In Vivo Activity.
Female nude mice (nu/nu), 3 to 5 weeks old, were obtained from Charles River Laboratories Inc. (Wilmington, MA) and used for human tumor xenografts. Tumors were established by subcutaneous injection of 1 × 106 HCT116 cells on the left rear flank of a nude mouse in a volume of 100 μl of sterile saline. Tumor dimensions were measured daily by using calipers, and tumor volumes were calculated by using the formula W2 × L × 0.5, where width (W) ≤ length (L). Tumor growth rate was analyzed based on slopes. The slopes were calculated by the difference between the final and initial tumor volumes and then divided by the number of treatment days: (Vn − V0)/n. V0 is the tumor volume (in mm3) when it first becomes apparent (20–100 mm3), V n is the tumor volume n days later, and n is the number of treatment days. To determine the largazole dose for efficacy studies, mice were treated with various doses of largazole (0, 1.5, 5, 15, and 50 mg/kg) by intraperitoneal injection, 4 h later tumors were harvested, and 50 mg of tumor tissue was sonicated in PhosphoSafe lysis buffer and used for immunoblot analysis probing with antiacetyl histone H3 antibody as described above. Mice were treated with the optimized dose of 5 mg/kg of largazole or solvent control every day until the tumor size in one dimension reached 15 mm (or the tumor volume exceeded 1000 mm3), and tumor tissue was harvested the day after final injection. Immunoblot analysis was performed as described above. Tumor tissue for immunohistochemistry was immediately kept in ice-cold 4% paraformaldehyde.
All studies were carried out under the protocol approved by the Institutional Animal Care and Use Committee at the University of Florida.
Xenograft tumors were fixed in 4% paraformaldehyde, embedded in paraffin, and sectioned. Sections were incubated with a 1:100 diluted caspase 3 primary antibody (Cell Signaling Technology) and a biotinylated 1:200 diluted secondary antibody (Cell Signaling Technology). After incubation with 100 to 400 μl of avidin-biotin peroxidase complex reagent, the slides were developed in 100 to 400 μl of 3,3′-diaminobenzidine. The slides were further counterstained with hematoxylin and dehydrated in ethanol and xylene. After the coverslips were mounted, the slides were scanned by using ScanScope CS (Aperio Technologies, Vista, CA), and the images were captured by using ImageScope (Aperio Technologies). The images were analyzed by using MetaMorph (Olympus, Tokyo, Japan). Caspase 3-positive stained cells were quantified on 12 randomly selected fields from two independent tumors.
Docking studies used a previously validated homology model of human HDAC1 (Wang et al., 2005). Structures for the largazole thiols were constructed based on a reported crystal structure of a synthetic intermediate (Seiser et al., 2008). The appropriate side chain atoms were added in Sybyl (Tripos, St. Louis, MO), then the structure was minimized. This process did not appreciably alter the conformation of the macrocyclic ring. In the case of the Tyr and His analogs, the isopropyl group of the Val side chain was replaced with the appropriate side chain in Sybyl before further minimization. Docking studies were carried out with AutoDock Vina (Trott and Olson, 2010), allowing rotation of the side chains but not within the macrocycle, and using an exhaustiveness value of 25. Figures of docked poses were prepared in PyMol (Schrödinger, New York, NY).
Largazole Exhibits Broad-Spectrum, Yet Differential, Activity in the National Cancer Institute's 60-Cell Line Screen.
To determine an appropriate cancer model system for initial biological studies, largazole was tested for antiproliferative activity in the NCI's cell viability screen using approximately 60 cancer cell lines derived from various solid and liquid tumors (Paull et al., 1995). Although we observed growth inhibition across all cancer cell lines, there was increased cytotoxicity in several cell lines that originated from melanoma, renal, and particularly colon cancer (Fig. 2A). We had observed previously potent activity of largazole during our early SAR studies using HCT116 colorectal carcinoma cells (Ying et al., 2008a). Consequently, we used this cell line for further biological studies.
Largazole Induces Dose-Dependent Cell Cycle Arrest and Apoptosis in HCT116 Colon Cancer Cells.
To rationalize the antiproliferative activity we determined largazole's effect on the cell cycle by using fluorescence-activated cell sorting-based DNA content analysis after 24 h of treatment. Whereas low concentrations of largazole induced G1 arrest (1–3.2 nM), higher concentrations (≥10 nM) shifted the population distribution significantly to an apparent G2/M arrest (Fig. 2B). The higher concentrations also caused apoptosis as evidenced by sub-G1 and sub-G2 populations. This result was consistent with the induction of caspase 3/7 activity (Fig. 2C), downstream mediators of programmed cell death.
SAR Study Identifies Critical and Tunable Elements in the Largazole Molecule.
When we replaced Val with Ala in the largazole structure we observed previously that largazole's activity is tunable at that position upon appropriate chemical modification (Ying et al., 2008a). We extended these studies by introducing aromatic (Phe, Tyr), acidic (Asp), or basic (His) amino acid residues at that position (Fig. 3A). All of the analogs exhibited reduced activity in HDAC1 and cell viability assays, but to different extents. The antiproliferative activity correlated with hyperacetylation of histone H3 (Lys9/14) and thus cellular HDAC inhibition in HCT116 cells (Fig. 3B). According to our docking studies to an HDAC1 homology model (Wang et al., 2005) with the corresponding thiols, the aromatic moieties (Phe, His, Tyr) were able to fill an adjacent pocket next to Glu98 of HDAC1 (Fig. 3, C–H). Although the Phe analog was the least active largazole analog of these three compounds, the His analog was slightly more active probably because of hydrogen bonding interactions with the ω-carboxylic acid unit of Glu98. The Tyr analog showed even better activity, presumably because of hydrogen bonding with the amide carbonyl of Glu98, which is not possible for the Phe analog (Supplementary Fig. S1). The Tyr analog had been prepared previously by others (Zeng et al., 2010); however, those studies used a different methodology rather than a validated HDAC1 homology model to rationalize activity, consequently leading to different docking results. In all of our structures, the mercapto group of each largazole thiol analog chelates the Zn2+ located in a tunnel in the active site of HDACs.
To probe the contribution of the two secondary amide hydrogens to largazole's HDAC inhibitory activity, we synthesized the two corresponding N-methylated largazoles and subjected them to HCT116 cells to test their effect on cell viability and cellular HDAC activity. Both compounds were 100- to 1000-fold less active (Fig. 3B). The crystal structure of the largazole macrocycle suggests hydrogen bonding of the Val-NH with the thiazoline substructure (Seiser et al., 2008), which would be lost on methylation, potentially leading to a conformational change and thereby reduced activity. Furthermore, in all structures obtained with secondary amides, the NH of the Gly-derived unit can form a hydrogen bond to Leu271 in the enzyme's active site (Fig. 3, C–H). The drastically reduced activity of the N-methylated analog at that position suggests that this interaction is crucial for HDAC inhibitory activity of largazole.
Expansion of the Phe-containing macrocycle resulted in loss of activity; we tested a dimeric macrocyclic product (Fig. 3A) that was obtained as a minor product during the macrocyclization reaction. The 400-fold drop in activity (Fig. 3B) indicates the importance of the macrocycle size of our HDAC inhibitor. Residual activities for the N-methylated analogs and the expanded macrocycle probably are caused by the presence of warhead (upon thioester hydrolysis) and the proper length of the spacer, retaining some (albeit not strong) ability to chelate the zinc ion in the active site of HDACs.
Transcriptional Changes Induced by Largazole.
Our next goal was to identify key transcriptional targets related to largazole's antiproliferative activity. To find optimal conditions for a genomic transcript analysis, we first established the optimal dose that can prevent deacetylation of histone H3 after 8 h. We verified by immunoblot that largazole (such as FK228 and in contrast to pan-inhibitor SAHA) does not inhibit cellular HDAC6 (class IIb), which is responsible for tubulin deacetylation (Fig. 4A). The endogenous cell cycle inhibitor p21 is one of the putative markers whose increased expression upon HDAC inhibitor treatment has been linked to the growth arrest-inducing ability of this class of agents (Wilson et al., 2006). We then used this marker to establish the optimal time point for genomewide transcriptional profiling, monitoring p21 expression in HCT116 cells upon largazole treatment (Supplementary Fig. S2A). For comparison, we also tested side by side SAHA and FK228, which we synthesized via a published route (Wen et al., 2008). Largazole and FK228 were approximately 100-fold more potent in these assays, and p21 transcript levels reached a maximum after 10 h (2.55- and 2.13-fold increases, respectively; Supplementary Fig. S2A). Therefore, we used RNA derived from 10-h-treated cells with all three drugs for a comparative global gene expression analysis. At the 2-fold threshold, largazole, SAHA, and FK228 (directly or indirectly) modulated transcript levels for hundreds of genes in either direction (Fig. 4B). Several up-regulated genes are cell cycle inhibitors other than p21 that can be linked to the drug-induced G1 growth arrest, such as p19, p15, and p57 (Supplementary Table S1). CDK6 and cyclin D1, encoding its regulatory subunit, were down-regulated, which may also contribute to the antiproliferative effects. Elevation of transcript levels of BCL2L11, which codes for a proapoptotic BCL2 protein family member, may in part explain the induction of apoptosis (Supplementary Table S1). Other down-regulated genes (Supplementary Table S1) include growth factor receptors (EGFR, HER-2, and MET) and IRS-1, which is downstream of insulin growth factor signaling and upstream of PI3K/AKT. The comparative transcriptional profiling revealed a close overlap of genes regulated by these three anticancer agents in HCT116 cancer cells (Supplementary Fig. S2B). Closer inspection of the relatively few genes tentatively categorized as uniquely regulated (Supplementary Fig. S2B) suggested that differences arise mainly from differences in the magnitude of fold change and may not reflect biological significance (Supplementary Table S2). Gene ontology analysis revealed that overrepresented functional classes of genes induced by largazole include those involved in chromatin assembly, negative regulation of cell cycle, and transcription. Using the same analysis, over-represented classes of genes down-regulated by largazole include genes with function in positive regulation of transcription, intracellular protein cascade, positive regulation of nucleic acid metabolism, RNA biosynthesis and metabolism, and epidermal growth factor receptor activity. We validated several regulated genes by Western blot analysis in a dose-response study after 24 h of treatment (Fig. 4C).
The Antiproliferative Effects of Largazole on Other Colon Cancer Cell Types Correlate with the Induction of Histone H3 Hyperacetylation through HDAC Inhibition.
To exclude the idea that the inhibition of HDACs in colon cancer cells by largazole is context dependent, we assessed histone H3 acetylation in another susceptible colon cancer cell line (HT29) and a less susceptible cell line (HCT15) based on the NCI-60 data (Fig. 2A). Whereas the acquisition of a malignant phenotype in HCT116 cells is associated with microsatellite instability caused by a defect in DNA mismatch repair because of mutL homolog 1 mutations, HT29 cells are characterized by adenomatous polyposis coli mutations (among others), leading to chromosomal instability. HCT15 cells exert both microsatellite and chromosomal instability caused by mutations in another DNA mismatch repair gene (mutS homolog 6) but also in the adenomatous polyposis coli gene (among others). As seen for HCT116 cells (Fig. 3B), in both HT29 and HCT15 cells the inhibition of histone H3 deacetylation correlates with the antiproliferative effects of largazole (Fig. 5, A and B). In addition, we observed a dose-dependent cell cycle arrest paralleling the effects of largazole on HCT116 cells (Fig. 2B). Concentrations near, but slightly below, the GI50 caused G1 arrest, whereas higher concentrations induced G2/M arrest (Fig. 5, C and D).
Aqueous, Plasma, Microsomal, and Cellular Stability of Largazole.
Stability studies are a prerequisite for predicting bioavailability in vivo. For largazole, those studies must also take into account the reactive species, largazole thiol, because this is the primary metabolite interacting with the HDACs. We monitored levels of largazole and largazole thiol after extraction from aqueous reaction mixtures with ethyl acetate followed by LC-MS analysis in the MRM mode. Largazole was remarkably stable in aqueous solution (Supplementary Fig. S3). However, in mouse serum largazole rapidly converted to the largazole thiol (Fig. 6A), suggesting that the thioester hydrolysis is mediated by plasma proteins. We also noticed the formation of a largazole thiol adduct by observing a peak at a retention time different from that of largazole thiol but with the same MRM transition as “free” largazole thiol (Fig. 6B), which initially signaled to us that the adduct formation was reversible. When largazole was incubated with mouse microsomes, we observed a similar trend in that >99% hydrolyzed within 5 min. However, it did not hydrolyze when the microsomes were inactivated (95°C, 5 min) before largazole exposure. This suggests that denatured proteins are unable to assist in the hydrolysis reaction. Active microsomal proteins also formed adducts with largazole thiol as observed for other proteins (Fig. 6), and the adduct was stable in the absence of NADPH. In the presence of NADPH, largazole thiol was metabolized by microsomes (approximately 19, 3, and 1% remaining after 5, 15, and 30 min, respectively; Supplementary Table S3). To establish the stability of largazole in a cellular context, we incubated largazole with whole-cell protein lysate derived from HCT116 cells and measured metabolite levels over time as described above. Largazole levels dropped concomitant with a rise in concentrations of a largazole thiol adduct, suggesting that protein-assisted hydrolysis and subsequent reversible protein binding had occurred (Fig. 6C). These experiments confirmed that when exposed to cellular proteins largazole hydrolyzes and generates the reactive largazole thiol.
In Vivo Activity and Efficacy of Largazole.
Acute toxicity studies in nu/nu mice indicated that largazole was well tolerated up to the highest concentration tested (50 mg/kg i.p.). The lack of toxicity in these studies prompted us to establish the appropriate largazole dose by examining the histone acetylation status in the HCT116 tumor xenograft as a biochemical marker of largazole activity rather than determining the maximum tolerated dose. Largazole induced hyperacetylation of histone H3 in the tumor after 4 h of treatment in a dose-response manner (Fig. 7A). These studies indicated that largazole at 5 mg/kg inhibited HDAC in vivo in the tumor and the inhibitor is sufficiently bioavailable. Having established the largazole dose required to cause a biochemical response, we subsequently carried out efficacy studies (daily intraperitoneally). We observed retarded HCT116 tumor growth in vivo (Fig. 7, B and C). Immunohistochemistry of control tumors and residual tumors from largazole-treated mice, staining for caspase 3, indicated that largazole induced apoptosis by approximately 10-fold in vivo (Fig. 7, D and E). To further compare our results with the mode of action of largazole in vitro, we analyzed protein extracts derived from control and largazole tumors. Consistent with the transcriptional effects observed in cell culture, largazole also induced levels of p15 and decreased levels of cyclin D1 and growth factor receptors such as HER-2 in vivo (Fig. 7F). Furthermore, IRS-1 was strongly down-regulated concomitant with inhibition of AKT (Thr308 and Ser473) phosphorylation, suggesting that the prosurvival PI3K/AKT pathway is inhibited in vivo (Fig. 7F). This is one potential explanation for the proapoptotic effect of largazole leading to caspase 3 activation. Thus, we confirmed that largazole's mechanisms of action in vitro and in vivo are at least partially overlapping.
The marine cyanobacterial metabolite largazole is among the most potent HDAC inhibitors found to date, displaying similar potency as FK228. Both compounds require metabolic activation through thioester hydrolysis or disulfide reduction, respectively. Largazole thiol possesses subnanomolar activity (Ying et al., 2008a) against recombinant HDAC1 and is more active than any other marine-derived HDAC inhibitor such as psammaplin (Piña et al., 2003) and azumamides (Nakao et al., 2006, 2008). In a systematic approach we determined that certain colon cancer cell lines are particularly susceptible to the cytotoxic effects of largazole and presented mechanistic, cellular SAR and in vivo efficacy results in HCT116 colon cancer. We showed that we can tune HDAC1 inhibitory and antiproliferative activities through systematic chemical mutagenesis of the Val residue. We have been able to decipher interactions between the molecule and HDAC1 enzyme by molecular docking. As we had originally suggested (Ying et al., 2008a), the HDAC1 pocket can tolerate various amino acid residues in this position without sudden loss of activity. Such a continuous spectrum of less potent compounds with differential activities may be exploited like SAHA, where more potent analogs proved to be too toxic (Marks and Breslow, 2007). Furthermore, we are now positioned to explore effects on isoform specificity. Our previous and present studies indicated that altering the macrocycle conformation, e.g., by inversion of C17 stereocenter (Ying et al., 2008a) or N-methylation, is detrimental to largazole's activity. In contrast, there is only limited SAR available for FK228 (Yurek-George et al., 2007).
Investigation of largazole's effect on the HCT116 transcriptome revealed close similarity to the effects of SAHA and FK228 tested in parallel. It seems likely that any major differences between these HDAC inhibitors would rather manifest themselves in other cell types, their effects on nonhistone targets, and their metabolic stability and bioavailability. Although the cell cycle inhibitor p21 is a standard biochemical marker for HDAC inhibition, we have shown that largazole also induces other cell cycle inhibitors and down-regulates CDK6 and its regulatory subunit, cyclin D1, which is overexpressed in many cancers and plays a critical role in tumor development (Tashiro et al., 2007). Thus largazole affects the cell cycle through various mechanisms, which is reflected in the dose-dependent effects on the cell cycle profile. At lower concentrations, modulation of genes involved in G1–S transition seems to dominate cell cycle behavior, whereas higher concentrations of largazole caused a pronounced G2/M arrest. These differences may be indicative of target-dependent dynamics or secondary effects and may be cell type dependent. Largazole also down-regulates several cancer-associated receptor tyrosine kinases. Specifically, levels of EGFR, HER-2, and MET, all of which are master regulators of many aberrantly activated signaling pathways in cancers, are substantially reduced in a dose-dependent manner. This effect on growth factor receptors is probably one major mechanism by which largazole exerts its potent antiproliferative activity. Numerous specific inhibitors of receptor tyrosine kinases, including EGFR (e.g., erlotinib), have been developed to treat different types of cancers. Resistance caused by mutations and cross-talk between signaling pathways, however, has caused a problem for the efficacy of these inhibitors, and, consequently, combination therapy with several receptor tyrosine kinase inhibitors is one promising clinical strategy (Baselga, 2006; Kling, 2006). Largazole regulates a subset of these receptors, which may be an advantage over single-agent therapy with specific growth factor receptor inhibitors. Another antiproliferative mechanism particularly related to the proapoptotic activity of largazole is the inhibition of AKT phosphorylation and thereby inhibition of a prosurvival pathway important to cancer cells. Other studies have shown that a subset of known HDAC inhibitors affects AKT independent of histone acetylation by disrupting HDAC–protein phosphatase 1 complexes (Chen et al., 2005). We have shown here that largazole strongly down-regulates IRS-1 in vitro and in vivo, which may be at least partly responsible for AKT inhibition. IRS-1 can activate PI3K and therefore AKT, but also acts as a transcriptional cofactor in promoters of cell cycle genes (Baserga, 2009). IRS-1 was found by chromatin immunoprecipitations to associate with the cyclin D1 promoter (Sun and Baserga, 2008), providing a link to why largazole may down-regulate cyclin D1. Furthermore, IRS-1 expression is often increased in tumors and IRS-1 plays a crucial role in cancer and transformation. The most effective microRNA (miR145) that targets the 3′ untranslated region of IRS-1 (among others) dramatically inhibits colon cancer cell growth (Shi et al., 2007).
Largazole's activity and therefore its therapeutic applicability are not limited to colon cancer cells with mutations in DNA mismatch repair genes (e.g., HCT116). HT29 cells, which display chromosomal instability, are similarly susceptible as HCT116 cells, and in all cases HDAC inhibition also correlated with largazole's antiproliferative activity even in less susceptible colon cancer cells such as HCT15.
We have shown that largazole is relatively stable in aqueous solution; however, its hydrolysis seems to be protein-assisted; it occurs in plasma and in the presence of active microsomal and cellular proteins. In comparison with FK228, which is reduced to redFK228 intracellularly, largazole thioester hydrolysis expectedly occurs faster to generate largazole thiol. However, in contrast to redFK228 (Furumai et al., 2002), largazole thiol is fairly stable, which is consistent with our observation that direct treatment of cancer cells with largazole thiol has strong antiproliferative activity. It has been reported that redFK228 and a related thiol analog exhibit much weaker activity caused by limited bioavailability and require delivery as a prodrug (Furumai et al., 2002; Yurek-George et al., 2007). It is noteworthy that largazole thiol is largely bound to proteins that may serve a dual function: adduct formation (via thioester or disulfide formation, or noncovalently) could “reprotect” largazole thiol (deprotection–reprotection or “trans-protection” strategy), but protein binding partners can also act as carrier proteins that release the active largazole thiol at the site of action, which (depending on the nature of the adduct) may occur hydrolytically, reductively, or by shifting the equilibrium toward formation of an HDAC–thiol complex because of the high enzyme affinity of the thiol. Although reversibility is suggested only indirectly because HDAC inhibition was observed in the tumor, reversible formation of a thiol adduct is an interesting strategy that may be explored for the development of other drugs as well. We have demonstrated that largazole thiol reaches the tumor and strongly inhibits HDACs responsible for histone H3 acetylation, validated the regulation of several proliferation and apoptosis-related target genes in vivo, and demonstrated in vivo efficacy.
In ongoing interactive chemical and biological studies we are systematically exploring the application of largazole and its derivatives to treat other types of cancer and study the detailed mechanisms of action downstream of HDAC inhibition. We are also extending our discovery beyond cancer and comprehensively applying largazole-based HDAC inhibitors to other disease indications where cellular reprogramming of transcription may be advantageous.
We thank Jason Li for assistance with bioinformatics.
This work was supported by the National Institutes of Health National Cancer Institute [Grant R01-CA138544].
Data Deposition: The microarray data have been deposited in the Gene Expression Omnibus database (accession no. GSE22061).
Article, publication date, and citation information can be found at http://jpet.aspetjournals.org.
- histone deacetylase
- cyclin-dependent kinase 6
- epidermal growth factor receptor
- ERBB2 (v-erb-b2 erythroblastic leukemia viral oncogene homolog 2)
- insulin receptor substrate 1
- met protooncogene (hepatocyte growth factor receptor)
- multiple reaction monitoring
- National Cancer Institute
- phosphoinositide 3-kinase
- suberoylanilide hydroxamic acid
- structure–activity relationship
- high-performance liquid chromatography-mass spectrometry
- declustering potential
- entrance potential
- collision energy
- collision cell exit potential
- collision cell entrance potential.
- Received July 1, 2010.
- Accepted August 23, 2010.
- Copyright © 2010 by the American Society for Pharmacology and Experimental Therapeutics