Human GBM U87 cell line was purchased from the American Type Culture Collection (ATCC). Cells were cultured in Dulbecco’s modified Eagle’s medium (DMEM) supplemented with 10% fetal bovine serum (FBS) and maintained in a 37°C, 5% CO₂ incubator. After cells were routinely passaged at 2- to 3-day intervals, experiments were carried out when the cells entered exponential growth phase.

The inhibition assays for HDACs were carried out according to standard protocols. Briefly, purified HDACs were incubated with the test compounds and a carboxyfluorescein-labeled peptide (1 μM) as substrate for 18 h at 25°C in a buffer consisting of 100 mM HEPES (pH 7.5), 1 mg/ml bovine serum albumin (BSA), 0.01% Triton X-100, 1% dimethyl sulfoxide (DMSO), and 25 mM KCl. The reaction was stopped by the addition of 45 μl of 100 mM HEPES (pH 7.5) and 0.08% sodium dodecyl sulfate (SDS). The substrate and the product were then separated electrophoretically using a LabChip 3000 system (Caliper Life Sciences, Hopkinton, MA) with blue laser excitation and green fluorescence detection.

Cells (5 × 10⁴) were seeded into 96-well plates and exposed to vehicle (PBS) or different concentrations of compounds for 72 h. Cell viability was measured using the 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) assay and was expressed as a ratio to the absorbance value at 570 nm of the control cells.

For apoptosis assays, 5 × 10⁵ U87 cells were seeded into six-well plates for 24 h and treated with a range of concentrations of DMC-HA for 36 h. Annexin V-FITC/PI apoptosis detection kit was used to perform the apoptotic assay following the manufacturer’s instructions (Beyotime, Jiangsu, China). After treatment, 1 × 10⁶ cells were washed using 1 ml binding buffer for three times, then centrifuged at 300×g and stained with 10 μl Annexin V-FITC solution at 37°C for 15 min. Before detecting, 5 μl PI solution was added to the samples, and the apoptotic cells were detected using flow cytometry (FACSCalibur BD, BD Biosciences, San Jose, CA, USA).

For cell cycle analysis, U87 cells (5 × 10⁵ cells/well) were seeded into six-well plates and treated with the indicated concentrations of DMC-HA for 24 h. Cells were detached, fixed in 70% ethanol in PBS (−20°C) overnight and resuspended in PBS supplemented with 100 µg/ml RNase and 50 µg/ml PI for 30 min. Cell cycle analysis was performed on a flow cytometry (FACSCalibur BD, BD Biosciences, San Jose, CA, USA).

Metabolic stability was assessed in the presence of human, mouse, and rat liver microsomes (XenoTech, Lenexa, KS, USA). All liquid dispenses and transfer steps were performed with the Freedom Evo automated liquid handler (Tecan US). Compound stock solutions were initially prepared in 100% DMSO and subsequently diluted in acetonitrile for the assay. The pH of the reactions was kept at 7.4 with potassium phosphate buffer. The reaction wells were prepared by adding microsomes to a well and allowed to warm to 37°C. Then, compound was added to each well. The reactions were stirred by adding cofactor reduced nicotinamide adenine dinucleotide phosphate (NADPH) to the reaction well containing microsomes and compounds. Negative controls received buffer only. Immediately after reactions started, 0-min aliquots were promptly collected and mixed in a separate well with ice cold acetonitrile to quench the reactions. The remainder of the reaction volume was incubated at 37°C with shaking. An additional aliquot was collected at 60 min after the start of the reaction and promptly quenched with ice cold acetonitrile. Samples were vortexed and centrifuged at 3,500 rpm for 10 min. The amount of compound in the supernatant was determined by liquid chromatography–tandem mass spectrometry (LC/MS/MS) (Thermo Fischer Scientific, San Jose, CA, USA), and the percent of parent compound remaining after 60 min was calculated. Results are the mean of each reaction triplicate, normalized to the internal standard, and expressed as the percent of compound remaining after the incubation time.

The animal studies were approved by the Institutional Animal Care and Used Committee. The pharmacokinetic study of DMC-HA was carried out with female SD mice. The compound (2.0 mM) was suspended in DMSO (10%) and Tween 80 (10%) in PBS (80%). A single dose of 1 mg/kg was administrated by intravenous injection (n = 3 mice), followed by blood collection via retro-orbital bleeding at 0.25, 0.5, 2.0, 4.0, 8.0, 12.0, and 24 h. Whole blood samples were added to a tube containing ethylenediaminetetraacetic acid (EDTA) and centrifuged. Plasma was collected and stored at −20°C. Compound standards were prepared with pooled plasma samples. Fifty microliters of thawed plasma and standards were added into wells of Ostro Pass-Through Sample Preparation Plate (Waters) for solid phase extraction. Acetonitrile (150 μl) containing internal standard was added and mixed by pipette. Next, clear sample solution was eluted by positive pressure processor under 60 psi for 5 min. The sample solution can be used directly for LC-MS/MS analysis.

U87 cells (5 × 10⁵ cells/ml) in six-well plates were incubated with DMC-HA and lysed in radioimmunoprecipitation assay (RIPA) buffer containing protease and phosphatase inhibitors. Protein contents in the lysates were determined by Bradford’s assay, and 50 μg was separated by SDS polyacrylamide gel electrophoresis (SDS-PAGE) electrophoresis. Proteins were transferred to membranes and probed with human anti-acetylated H3 and anti-β actin antibodies as a loading control. Membranes were then labeled with the appropriate horseradish peroxidase (HRP)-conjugated immunoglobulin G (IgG), and protein bands were visualized using enhanced chemiluminescence. (Bio-Rad, Hercules, CA, USA).

Six-week-old BALB/c nude mice were purchased from Shanghai Experimental Animal Center of Chinese Academy of Science (Shanghai, China). All animal protocols were implemented according to the guidelines of the Association for Assessment and Accreditation of Laboratory Animal Care. U87 cells (3 × 10⁶) were subcutaneously implanted into the right flank of mice. After 6 days of tumor growth, tumors grew to approximately 100 mm³, and mice were randomly divided into three groups with nine animals in each group. Mice in two groups were intraperitoneally injected with 20 mg/kg vorinostat or DMC-HA, respectively, every day for a total of 13 days. The control group was intraperitoneally injected an equal volume of DMSO. Body weights and tumor volumes were measured every 2 days. After treatment, all of the mice were sacrificed, and the tumors were harvested, weighted, and photographed. The tumor volume was calculated according the formula: tumor volume = (length × width²)/2.

All tests were performed using SPSS Graduate Pack 11.0 statistical software (SPSS, Chicago, IL, USA). Descriptive statistics, including the mean and SE, in addition to one-way ANOVAs, were used to determine significant differences. *p < 0.05 and **p < 0.01 were considered statistically significant.

We started our project with reference to the well-established cap-linker-metal binding group pharmacophore model for HDAC inhibitors. This model represented by a capping group is able to interact with the rim of the catalytic tunnel of the enzyme, opposite to a zinc-binding group (ZBG) at the bottom of the catalytic cavity, and a carbon linker connecting the two parts (27). To date, numerous studies have established that hydroxamic acid is a powerful Zn-chelating group; three Food and Drug Administration (FDA)-approved HDAC inhibitors (vorinostat, panobinostat, and belinostat; Figure 1A) also contain this ZBG (28). The predicted docking pose of curcumin (Figure 1B) suggested that one of the phenyl of curcumin was positioned next to the zinc cation; thus, we envisaged that insertion of hydroxamic acid group into one of the benzene rings of curcumin may facilitate its occupation of the catalytic site and enhance the HDAC binding affinity, leaving the second phenyl group toward the solvent space (capping region). Finally, the structure of curcumin-based HDAC inhibitor is given in Figure 1, namely, DMC-HA.

To date, 18 HDAC enzymes have been identified and grouped into four classes. Class I (HDACs 1–3 and 8), class II (HDACs 4, 5, 7, and 9 as class IIa and HDACs 6 and 10 as class IIb), and class IV (HDAC 11) are Zn-dependent isozymes, while class III isozymes (Sirt1-7) were quite different from the other ones as they utilize NAD as a cofactor. We systematically investigated the in vitro enzymic inhibitory activity of DMC-HA on different class of HDAC isoforms. The inhibitory activity against HDACs 1, 3, 4–6, 8, 10, and 11 subtypes and SIRT1 of DMC-HA was measured, and the results are shown in Table 1. Vorinostat is used as a control compound. DMC-HA exhibited IC₅₀ values of 0.51, 1.67, 0.38, 2.93, and 8.39 μM against HDACs 1, 3, 6, 8, and 10, respectively, while the IC₅₀ values on the four other subtypes exceeded 10 μM. Although both vorinostat and DMC-HA are inert to HDAC4 and HDAC5, it has been reported that class IIa (including HDACs 4, 5, 7, and 9) enzymes are inefficient on histone substrates; class I HDACs, especially HDACs 1–3, and HDAC 6 (class IIb) are considered key targets for cancer treatment (29). Overall, the current biochemical data showed that DMC-HA is a potent inhibitor of class I and class IIb isoforms.

As DMC-HA was identified as a potent novel inhibitor for HDACs, its antiproliferative activities on a panel of cancer cell lines and normal cells were evaluated, including four glioma cell lines, and the results are shown in Table 2. Vorinostat and curcumin were also tested as positive controls. Curcumin had moderate antiproliferative activities with IC₅₀ values at the range of 16.4–49.2 μM against all eight cell lines, whereas its hydroxamic acid derivative, DMC-HA, exhibited more potent activities with IC₅₀ values below 5.3 μM against these cell lines, suggesting a remarkable potency increase. In particular, DMC-HA is a potent compound against four glioma cell lines, with IC₅₀ of 2.99 in U251, 2.78 μM in U87, 4.62 μM in SHG44, and 3.38 μM in LN229 cell line, which were comparable with that of positive control vorinostat. More importantly, DMC-HA exhibited low toxicity to human umbilical vein endothelial cells (HUVECs), suggesting the safety of DMC-HA.

Next, DMC-HA-induced acetylation of histone H3 in living cells was analyzed using Western blot method. We compared the inhibition capacities of DMC-HA and vorinostat against HDACs in U87 cells first. After the incubation with various concentrations of DMC-HA or vorinostat, acetylated H3 levels in U87 cells were tested. As shown in Figure 2A, DMC-HA and vorinostat dose dependently increased the acetylation of histone H3 after the cells were treated for 24 h. However, it seems that vorinostat more significantly elevated the acetylation of histone H3 compared to that of DMC-HA at the same concentrations. Then, the acetylation of histone H3 after treatment with 5 μM DMC-HA for different times were detected, and the results showed that it increased in a time-dependent manner (Figure 2B).

Inhibition of HDACs results in anticancer effect through various mechanisms, like cell cycle arrest, reduced proliferation, and induction of apoptosis (30). Thus, the cell cycle profile of the human GBM cell line U87 treated with DMC-HA was analyzed by using flow cytometry. As reported in Figure 3, after incubation with different concentrations of DMC-HA for 36 h, DNA content analysis of cells revealed a strong increase in G2/M phase population and an associated decrease in G0/G1 population. Although in control cells, only 4.6% of cells were in G2/M phase, 14.5% of cells in G2/M phase were found after treatment with 2.5 μM DMC-HA. Besides, an increase in sub-G0/G1 population, an indication of cell apoptosis, was also observed after the incubation of DMC-HA, suggesting that DMC-HA could promote apoptosis.

To further investigate the apoptotic effect of DMC-HA on GBM, we treated U87 cells with increasing concentrations of DMC-HA. After 48 h, the morphological changes of DMC-HA-treated U87 cells were observed, such as cell shrinking, rounding, and detachment (Figure 4A). In addition, an increase in lactate dehydrogenase (LDH) release was correlated with increasing concentrations of DMC-HA after 48 h of treatment (Figure 4B), suggesting that DMC-HA induces cell death in a dose-dependent manner in U87 cells. To further confirm whether DMC-HA could induce apoptosis in U87 cells, cells were stained with Annexin V and PI after treatment with various concentrations of DMC-HA. Flow cytometry analysis of the cells is shown in Figure 4C; DMC-HA treatment induced an increase in both early- and late-stage apoptosis of U87 cells, and treatment with 10 μM DMC-HA resulted in an increase in total apoptotic cells from 3.3% up to 90.0% (Figure 4D). Together, these results indicated that DMC-HA can trigger apoptosis of U87 cells.

Structures that are active in vitro may not be potent in vivo due to their susceptibility to metabolism in the body. We then thoroughly examined the liver microsomal metabolic stability of DMC-HA across three different species (Table 3). Curcumin is indeed metabolically labile in mice, rat, and human with short half-life and high clearance rate. However, DMC-HA is relatively more metabolically stable; the half-life in three species is two to five times longer than those of curcumin, suggesting that the introduction of the hydroxamic acid group might block the metabolically labile sites and increase the metabolically stability. In addition, the half-life of DMC-HA in human microsome is generally two to three times longer than those in mice and rats. Collectively, the in vitro liver microsomal stability of DMC-HA is favorable, guaranteeing its further in vivo studies.

As DMC-HA exhibited adequate metabolic stability in microsomal stability assay, we decided to evaluate its in vivo pharmacokinetic profiles and assess the relationship between in vitro and in vivo pharmacokinetic data. The preliminary in vivo pharmacokinetic study for DMC-HA was performed in mice. SD mice were intravenously injected with 1 mg/kg DMC-HA or orally administrated with 10 mg/kg DMC-HA. Plasma samples were collected and analyzed with LC-MS/MS. Following p.o. administration, DMC-HA reached its peak concentration of about 1.0 μg/ml near 40 min, and moderate plasma clearance (8.7 L/h/kg) and half-life (4.1 h) were observed (Table 4). In addition, DMC-HA displayed acceptable oral bioavailability (40.2%).

Discovery of drugs for the treatment of GBM encounters formidable challenges to their ability to cross the blood–brain barrier (BBB) (31). Thus, we further investigated the distribution of DMC-HA to identify whether it has the ability to cross the BBB. The results in Figure 5 show that although the brain concentrations of DMC-HA were not high relative to the plasma exposure (brain-to-plasma ratios of 0.08–0.23), it can indeed penetrate the BBB and distribute from plasma into brain tissue. In addition, stable and durable distribution of DMC-HA in the brain was observed after p.o. administration. Nevertheless, optimization for the balance among potency, selectivity, metabolic stability, and brain permeability in a single molecule was challenging. These preliminary results provide valuable information for further structure optimization of DMC-HA and a guidance for its following in vivo studies.

Given the promising in vitro properties and good pharmacokinetic profiles of DMC-HA, we evaluated its in vivo behavior in U87 tumor xenograft model. U87 cells were subcutaneously inoculated into the right flank of mice. The mice were randomly divided into three groups (n = 9); the positive control vorinostat and DMC-HA were administered by the oral route once daily at the dose of 20 mg/kg. The resulting tumors were excised from the animals after treatment (Figures 6A, B). Both compounds showed no observable toxicity during the administration period and had no effects on the body weight (Figure 6D). Vorinostat at the dose of 20 mg/kg showed moderate inhibition on the tumor volume (24.7%, p = 0.04 vs. control) (Figure 6C). Interestingly, DMC-HA was much more active than vorinostat at the same dose, which inhibited the tumor volume with a value of 52.1% (p = 0.00083). Taken together, these data indicated that DMC-HA was efficacious in inhibiting the growth of GBM in vivo and deserved further evaluations.