α-Keto Acid Metabolites of Naturally Occurring Organoselenium Compounds as Inhibitors of Histone Deacetylase in Human Prostate Cancer Cells

Prostate cancer is the most common cancer in men in the United States. Although there is an unequal burden by race and ethnicity, the mortality rate is moderate. In 2009, >192,000 American men will be diagnosed with prostate cancer and 27,360 men will die from this disease (1). Because most treatments can have significant side effects, development of chemopreventive agents and strategies for the prevention of prostate cancer are highly desirable. Epidemiologic studies and clinical intervention trials have shown a protective role of selenium compounds against prostate cancer (2–7). A clinical trial by Clark and his colleagues, as well as follow-up studies, provided support for the protective role of selenium-containing compounds against the progression of human prostate cancer (6, 7). Recently, however, the Selenium and Vitamin E Cancer Trial (SELECT) was halted due to an apparent lack of efficacy for supplemental vitamin E and organoselenium, either alone or in combination (8, 9).

An important caveat to translational studies is to understand the precise chemical nature of both inorganic and organoselenium compound(s) because toxicity and therapeutic efficacy depend greatly on the chemical form and not on the amount of elemental selenium. The SELECT study used purified selenomethionine, whereas some earlier diet studies used selenium-enriched yeast, which like seleniferous plants such as garlic, onions, and broccoli, also contains the naturally occurring organoselenium compound, Se-methyl-l-selenocysteine (MSC; refs. 10–12). The anticancer potential of MSC has been reported in human breast cancer cells (13–16), mouse mammary tumor cell lines (17, 18), and a mammary tumor model in rats (19). Recent laboratory studies (20) revealed a possible use of MSC for the prevention of prostate cancer by showing that MSC can inhibit the growth of human prostate cancer cells in a xenograft mouse model.

The chemopreventive efficacy of MSC and other organoselenium compounds has been suggested to result from in situ generation of methylselenol (CH₃SeH) by β-lyases (21). Methylselenol can also be generated by intracellular reduction of dimethyldiselenide and methylseleninic acid by endogenous glutathione. Thus, methylselenol has been suggested to play a critical role in chemoprevention and the anticancer properties associated with selenium supplementation through its ability to alter cell signaling pathways and induce cellular apoptosis (17).

In order to generate methylselenol in situ using naturally occurring organoselenium compounds, MSC and/or selenomethionine must undergo a β-elimination and a γ-elimination reaction, respectively (22). β-Lyases and γ-lyases are pyridoxal 5′-phosphate (PLP)–dependent enzymes. Curiously, many aminotransferases can also catalyze β-elimination reactions, particularly with cysteine S-conjugates that possess a good electronegative-leaving group. For these enzymes, aminotransferase reactions often compete with elimination reactions. A full transamination reaction requires the concomitant presence of an α-keto acid substrate or a steady supply of PLP because pyridoxamine 5′-phosphate, the coenzymic product in a half-transamination reaction, unlike PLP, cannot catalyze the β-lyase reaction (23, 24). Selenocysteine Se-conjugates, in particular MSC, have been shown to be 5-fold to 10-fold better aminotransferase substrates of rat kidney glutamine transaminase K (GTK) than of the corresponding cysteine S-conjugates (25). GTK is an aminotransferase that is widely distributed in rat tissues and the rat enzyme has broad specificity toward glutamine, methionine and other sulfur and selenium amino acids, aromatic amino acids, and the corresponding α-keto acids (26). Human GTK has also been shown to have broad substrate specificity (27). Of interest to the current work, MSC was found to be both an aminotransferase and β-lyase substrate of human GTK (27). Because of the weaker C—Se bond compared with the C—S bond and/or the more facile abstraction of the β proton, selenocysteine Se-conjugates exhibit greater aminotransferase and β-lyase reactivity than the corresponding cysteine S-conjugates (25, 28). These characteristics of MSC led us to question whether methylselenol was the only critical metabolite to explain the chemopreventive activity of organoselenium compounds >in a physiologic environment. Accordingly, under appropriate enzymatic conditions of regenerating PLP or in the presence of α-keto acid substrates, MSC should undergo a transamination reaction that competes with β-elimination. Thus, in addition to in situ formation of methylselenol through a β-elimination reaction, the deaminated product of MSC, β-methylselenopyruvate (MSP), should also be formed (Fig. 1).

MSP, as well as the α-keto acid product of selenomethionine, i.e., α-keto-γ-methylselenobutyrate (KMSB) resembles butyrate, an inhibitor of histone deacetylase (HDAC). HDAC inhibitors are showing promise for the treatment of several human cancers, and several mechanisms have been proposed (29–32). To date, five classes of HDAC inhibitors have been identified (a) short-chain fatty acids, for example, butyric acid; (b) hydroxamic acids, such as suberoylanilide hydroxamic acid; (c) electrophilic ketones that include trifluoromethyl α-ketones and α-ketoamides; (d) aminobenzamides, e.g., MS-275 and CI-994; and (e) natural cyclic peptides such as apicidin. The structural multiplicity of HDAC inhibitors reflects both the diversity of the substrates for HDAC and the heterogeneity of tumor cell phenotypes. An intriguing possibility for the oxidatively deaminated products of either MSC or selenomethionine is the presence of a selenium moiety capable of disrupting the charge relay system in the HDAC pocket by coordinating with the zinc cofactor (33, 34).

This study shows the previously unrecognized effects of MSC and selenomethionine as prodrug inhibitors of HDACs. We provide information, for the first time, that the α-keto acid metabolites of MSC and selenomethionine, in addition to methylselenol derived from β-lyase and γ-lyase reactions, may be potential direct-acting metabolites of organoselenium in the chemopreventive activity for prostate cancer. These studies provide a new understanding of the mechanisms by which naturally occurring organoselenium may decrease the progression of human prostate cancer.

Se-Methylseleno-l-cysteine (MSC), l-selenomethionine (selenomethionine), L-phenylalanine, ammediol [2-amino-2-methyl-1,3-propanediol], metaphosphoric acid (MPA), trichostatin A (TSA), sodium butyrate, and sodium α-keto-γ-methylthiobutyrate (KMB) were purchased from Sigma-Aldrich. A synthetic organoselenium compound, p-XSC [1,4-phenylene bis (methylene)selenocyanate] as well as the cysteinylated and the glutathionylated derivatives were synthesized as reported previously (35–37) and was a gift from Dr. Karam El-Bayoumy, Penn State, Milton S. Hershey Medical Center, Hershey, PA. Recombinant human GTK (rhGTK; also referred to as human kynurenine aminotransferase I) was obtained by the method of Han et al. (38) and generously supplied (16 mg/mL, 18 units/mg in the standard phenylalanine-KMB assay; see below) by Dr. Jianyong Li, Department of Biochemistry, Virginia Tech., Blacksburg, VA. Crotalus adamanteus l-amino acid oxidase (LAAO; 7.3 units/mg) and bovine liver catalase (200,000 units/mL) were purchased from Sigma Chemical Company.

Androgen-responsive LNCaP and androgen-independent PC-3 and DU145 cells were obtained from the American Type Culture Collection and the androgen-independent clone of LNCaP cells, LNCaP C4-2, was a generous gift from Dr. Warren D.W. Heston (The Lerner Research Institute, the Cleveland Clinic Foundation, Cleveland, OH). LNCaP, LNCaP C4-2, and PC-3 cells were cultured in phenol red-free RPMI 1640 (Life Technologies/Invitrogen) supplemented with 5% (v/v) fetal bovine serum (FBS; Mediatech, Cellgro) and 1× nonessential amino acid solution (Mediatech, Cellgro). DU145 cells were cultured in DMEM (Life Technologies/Invitrogen) supplemented with 5% (v/v) FBS, 1× nonessential amino acid solution, and 1 mmol/L of pyruvate. All four cell lines were seeded 48 to 72 h before the experiments (1 × 10⁶) into 100 mm culture dishes and cultured in a humidified incubator at 37 °C and 5% CO₂. After treatments (10-50 μmol/L of MSP or KMSB; 100-200 μmol/L MSC or selenomethionine; 2.5 mmol/L sodium butyrate or 20 μmol/L TSA) for 5 and 24 h, cells were washed twice with ice-cold PBS and harvested for Western blot analysis into radioimmunoprecipitation assay buffer (150 mmol/L NaCl, 1.0% Triton X-100, 0.5% sodium deoxycholate, 0.1% SDS, and 50 mmol/L Tris; pH 8.0) including protease inhibitors, 200 μmol/L of 4-(2-aminoethyl) benzenesulfonyl fluoride hydrochloride, 100 μmol/L of leupeptin, 800 nmol/L of aprotinin, and 1 μg/mL of pepstatin A. Cells were also harvested into PBS and their nuclear and cytosolic fractions were separated using NE-PER nuclear and cytoplasmic extraction reagents (Pierce Chemical, Co.) for HDAC assay. All samples were stored at −80°C until use.

HDAC activity was determined in nuclear and cytosolic extracts using the fluorometric HDAC activity assay kit (BioVision) according to the instructions of the manufacturer. This assay is based on the Fluor de Lys (fluorogenic histone deacetylase lysyl) substrate and developer combination. Incubations were done at 37°C for 30 min with nuclear or cytosolic extracts from human prostate cancer cells (10-50 μg total protein) or with HeLa nuclear extract (supplied with the kit), and the reaction was initiated by the addition of HDAC substrate [Boc-Lys(Ac)-AMC]. Both androgen-responsive (LNCaP) and androgen-independent (C4-2, PC3, and DU145) prostate cancer cells were used for HDAC activity measurements. After 30 min, lysine developer was added and the mixture was incubated at room temperature for 30 min. Fluorescence was measured in a SpectraMax Gemini XS Microplate (Molecular Devices Corporation) using an excitation wavelength of 365 nm and an emission wavelength of 450 nm. HDAC activity was expressed as the relative ratio of HDAC activity in control (basal level) to that in cells treated with MSC, selenomethionine, and their α-keto acid products after calculating the relative fluorescence units per microgram of protein.

Protein concentrations were determined using the BCA protein assay kit with bovine serum albumin as a standard (Pierce Chemical). For Western blot analysis, 20 to 100 μg of total soluble proteins were separated by one-dimensional SDS-PAGE using 10% or 12% (w/v) gels under reducing conditions. Proteins were transferred to a polyvinylidene fluoride membrane (Millipore) and the membrane was blocked for 1 h with 10% nonfat milk solution (Carnation), probed with one of two primary antibody solutions for acetylated histone H3 and histone H3 (Santa Cruz Biotechnology) for 2 h at room temperature or overnight at 4°C. The membrane was then probed with secondary antibody solution (goat anti-rabbit IgG or goat anti-mouse IgG, and peroxidase conjugated; Pierce Chemical) for 1 h at room temperature. Proteins were visualized by developing the membrane using chemiluminescence reagents (Pierce Chemical). Each membrane was stripped and reprobed with anti–β-actin antibody (Santa Cruz Biotechnology). Equivalent sample loading was further confirmed using Ponceau S staining (Boston BioProducts).

The GTK assay in whole cell extracts was done according to the published procedure (27), measuring transamination between l-phenylalanine and KMB. The reaction mixture consisted of 20 mmol/L of l-phenylalanine, 5 mmol/L of KMB, 100 mmol/L of ammediol-HCl buffer (pH 9.0), and homogenates from prostate cancer cells. After incubation at 37°C for 30 min, the reaction was stopped by adding 0.15 mL of 1 mol/L NaOH and absorbance of phenylpyruvate-enol was measured at 320 nm wavelength (ϵ = 16,000 mol/L/cm). The blank reaction lacked KMB. For determining the substrate specificity of GTK for MSC and selenomethionine, a stock solution of rhGTK was prepared (57.6 mU/mg protein) in 15 mmol/L of potassium phosphate buffer (pH 6.8) containing 20% glycerol and stored at 4°C until use. The reaction mixture consisted of 10 mmol/L of MSC or selenomethionine, 5 mmol/L of KMB, 100 mmol/L of sodium phosphate buffer (pH 7.4), and 3.2 μg of purified rhGTK. The change in buffer and pH (from 9.0 to 7.4) provides comparable GTK activity but under physiologic conditions. In addition, when MSC and selenomethionine were used as amino acid substrates, and KMB as a cosubstrate, concentrations of the corresponding transamination products, i.e., MSP, KMSB, and methionine, were measured by high-performance liquid chromatography (HPLC) using CoulArray detection (see below).

To synthesize the α-keto acid metabolites of MSC and selenomethionine, each of the organoselenium compounds was incubated for 2 h with LAAO. The reaction mixture consisted of 5 mmol/L of MSC or selenomethionine, 100 mmol/L of potassium phosphate buffer (pH 7.4), 100 unit of catalase, and 0.1 units of LAAO in a total volume of 0.5 mL. The reaction was conducted at 37°C for 1 h and deproteinized by either the addition of 25% of MPA or passing the reaction mixture through a YM-10 Microcon Centrifugal Filter (10,000 nominal molecular weight cutoff limit; Millipore Corporation) prior to measuring the corresponding α-keto acid products of MSC and selenomethionine by HPLC.

The α-keto acids generated enzymatically from MSC and selenomethionine, i.e., MSP and KMSB, respectively, were analyzed by HPLC with electrochemical detection. When transamination reactions using rhGTK were done with MSC and selenomethionine, the α-keto acid cosubstrate used was KMB. After the transamination reaction, the resulting amino acid formed from KMB is methionine, which was also detected by HPLC using CoulArray detection.

Because MSC, selenomethionine, MSP, KMSB, methionine, and KMB are redox-active compounds, they are easily determined by HPLC with CoulArray detection without the need for prior derivatization. After reaction with rhGTK or LAAO, the reaction was stopped by the addition of 25% (w/v) MPA to yield a 5% (w/v) MPA solution. A Rheodyne injection valve with a 5 μL sample loop was used to manually introduce samples directly onto a Bio-Sil ODS-5S, 5 μm particle size, 4.0 × 250 mm, C18 column (Bio-Rad, Life Science Research Group). Samples were eluted with a mobile phase consisting of 50 mmol/L of NaH₂PO₄, 3% (v/v) acetonitrile, 1 mmol/L of octanesulfonic acid (pH 2.6) at a flow rate of 1 mL/min. PEEK (polyetheretherketone) tubing was used throughout the HPLC system, and 0.2 μ PEEK filters were placed pre-column and post-column to protect both column and flow cells, respectively, from any particulate matter. The eight-channel CoulArray detectors (ESA, Inc.) were set at 175, 250, 325, 400, 475, 550, 700, and 800 mV, respectively.

Negative electrospray ionization spectra were determined on an Applied Biosystems 4000 Q Trap (Triple quadrupole) mass spectrometer for compounds MSP and MSKB. Prior to the mass infusion analysis, the synthetic mixture was desalted using a 1 mL, 30 mg Oasis MAX SPE column from Waters. The cartridges were conditioned with 1 mL of methanol, and then equilibrated with 1 mL ofwater, loaded with 50 μL of the synthetic mixture added with 200 μL of H₂O and 30 μL of 28% NH₄OH. The column was washed with 1 mL of H₂O and eluted with 2 mL of 2% formic acid in methanol. The eluant was evaporated to dryness with a SpeedVac and the sample reconstituted with 150 μL of 50:50 acetonitrile/0.1% formic acid. For compound MSP, the negative peaks observed at m/z 178, 179, 181, and 183 reflect the isotopic distribution of Se (Fig. 2), which is consistent with the natural abundance of selenium isotopes (Se⁷⁴, Se⁷⁶, Se⁷⁸, Se⁷⁷, Se⁷⁹, Se⁸⁰, and Se⁸²), thus indicating the presence of a single selenium atom in a molecular formula of C₄H₆O₃Se for MSP. Similarly, the peaks observed at m/z 193, 195, and 197 (C₅H₈O₃Se) are the negative molecular ions of KMSB corresponding to Se⁷⁸, Se⁸⁰, and Se⁸².

The three-dimensional coordinates of bacterial Aquifex aeolicus HDAC homologue catalytic domain were taken from the crystal structure data available in the Protein Data Bank 1C3R (TSA/HDAC homologue; refs. 39, 40). The model was energetically refined in the internal coordinate space with Molsoft ICM v3.5-1p. The histidyl moieties at HDAC sites 131 and 132 were selected to be in the δ- protonated state, which was critical for demonstrating proper docking of hydroxamic acid–based inhibitors (40).

The molecular system is described in terms of internal coordinate variables, using the electrostatically driven Monte Carlo method with the Empirical Conformational Energy Program for Peptides/3 force field (41), with distance-dependent dielectric constants for energy calculations as implemented in ICM (42). The method used in our studies is an extension of the biased probability Monte Carlo with minimization procedure for global energy optimization (43, 44). The biased probability Monte Carlo global energy optimization method consisted of the following steps: (a) a random conformation change of the free variables according to a predefined continuous probability distribution (42, 43); (b) local energy minimization of analytical differentiable terms; (c) calculation of the complete energy including nondifferentiable terms such as entropy and solvation energy; (d) acceptance or rejection of the total energy based on the Metropolis criterion (45) and return to step (a).

The internal coordinate mechanics (ICM) structure-based virtual screening method was used (46) to simulate docking of each selenoketo acid. The receptor is represented by six grid potential maps accounting for hydrophobicity, van der Waals interactions, hydrogen-bonding and electrostatic potential. The ligand is considered fully flexible in the field of the receptor. To ensure the convergence of the biased probability Monte Carlo–based global energy minimization, each ligand was docked thrice into the receptor, and the best scoring configuration for each compound was attained (Molsoft ICM v3.5-1p).

Results were statistically analyzed by a one-way ANOVA and Dunnett's comparison to test the significance of differences in measured variables between control and experimental groups in Minitab (release 13.1). All enzyme assays were performed in triplicate. Statistically significant differences are denoted at P < 0.05 or P ≤ 0.01.

Based on previous findings that selenocysteine Se-conjugates are GTK and LAAO substrates (24, 26), we sought to prepare the corresponding α-keto acid derivatives of MSC and selenomethionine as depicted in Fig. 1. Mass spectral analysis of the reaction products formed during 2 hours of incubation with LAAO at 37°C revealed α-keto acid products of MSC and selenomethionine, i.e., MSP and KMSB, respectively (Fig. 2), both compounds exhibited patterns containing the naturally abundant isotopes of selenium. As MSC and selenomethionine were stoichiometrically metabolized into α-keto acid products via the action of LAAO (data not shown), we used this procedure to separate and quantify MSP and KMSB using HPLC with electrochemical detection.

For the initial purification and characterization of the reaction products from rhGTK and prostate cancer cell lysates, reaction mixtures were precipitated with 5% MPA, and were subjected to HPLC analysis using a C18 column. Compounds were eluted with phosphate buffer containing 3% acetonitrile. The elution pattern was monitored using coulometric detection. Figure 3 illustrates the peak elution times and oxidation patterns of a standard mixture containing MSC, MSP, methionine, KMB, selenomethionine, and KMSB and having retention times at 3.6, 4.0, 4.5, 5.2, 5.4, and 6.6 minutes, respectively. Ascorbic acid (retention time 3.2 minutes; 175 mV) was incorporated in the mixture as a redox protectant.

Commandeur et al. (25) showed that rat kidney GTK catalyzes both a β-elimination reaction and a transamination reaction with MSC. The transamination product was not characterized in that study. Recently, Cooper et al. (26) showed that rhGTK catalyzes a similar reaction with MSC but not to any appreciable extent with selenomethionine. Here, we confirm that the α-keto acid product of MSC is MSP, which structurally resembles the known HDAC inhibitor, butyrate. We hypothesized that prostate cancer cells possess GTK or other aminotransferases capable of generating MSP. Androgen-responsive (LNCaP) and androgen-independent (C4-2, PC-3, and DU145) cells were tested for GTK activity using the standard phenylalanine-KMB assay mixture. Table 1 shows the presence of GTK in human prostate cancer cells and compares this activity with that present in rat kidney.

Subsequent measurements of human prostate cancer cell extracts using MSC rather than phenylalanine as the amino acid substrate and KMB as the α-keto acid cosubstrate revealed the formation of MSP (Fig. 4). In this experiment, MSP was generated at least in part from the endogenous activity of GTK. Specific activities for LNCaP, LNCaP C4-2, PC-3, and DU145 were 1.05, 0.56, 0.72, and 0.82 μmol of MSP formed per hour/milligram of protein, respectively.

Because GTK has an unusual “crown” of aromatic amino acid residues in the binding pocket (47), and can accommodate a variety of large amino acid substrates, e.g., 5-S-l-cysteinyl-l-DOPA and 5-S-l-cysteinyldopamine (27), we also considered the possibility that the cysteinylated or glutathionylated derivative of the synthetic organoselenium compound, p-XSC, may also undergo transamination. However, neither rhGTK nor the prostate cancer cell extracts revealed the formation of any corresponding keto acid product (data not shown).

Naturally occurring (MSC, selenomethionine) organoselenium compounds were added to cell culture medium and the expressed level of acetylated histone 3, as well as nonacetylated histone 3 (as loading control), were examined by immunoblot analyses of whole cell lysates (Fig. 4A). Compared with controls, MSC-treated cells expressed increased levels of acetylated histone 3 after 5 hours of incubation in LNCaP and LNCaP C4-2 cells, but slightly lower levels in PC-3 cell lysates. The finding that LNCaP cells showed a more marked accumulation of acetylated histone 3 in the presence of MSC at both 0.05 and 0.2 mmol/L concentrations after 5 hours of incubation may be reflective of the higher specific activity of GTK for MSC in LNCaP compared with LNCaP C4-2 and PC3 cells (see Fig. 3). However, the expression of acetylated histone 3 decreased after 24 hours of incubation, suggesting a rapid, transient response. By contrast, incubation of cells for up to 48 hours with selenomethionine did not show an increase in acetylated-histone 3 in any of the cell lines tested, further demonstrating that it is not an active substrate for transamination in human prostate cancer cells. Treatment of cells with the α-keto acid forms of MSC and selenomethionine, i.e., MSP and KMSB, resulted in the elevation of acetylated histone 3 in all cell lines within 5 hours at 50 μmol/L concentration, supporting our hypothesis that the increased level of acetylated histone 3 following MSC treatment is due to its metabolite, MSP (PC-3 cells did not survive with 50 μmol/L of MSP or KMSB concentrations after 24 hours of incubation). In addition, direct incubation of cells with p-XSC for 24 hours did not show the accumulation of acetylated-histone H3 in prostate cancer cells (data not shown). Thus, neither the selenoamino acids, MSC and selenomethionine, nor the selenoketo acids, MSP and KMSB, affect total histone 3 levels and only the selenoketo acids increase the ratio of acetylated to total histone 3.

As shown in Fig. 4B, upper panel, MSC had no direct HDAC-inhibitory effect in nuclear fractions of the four prostate cancer cell lines, despite the observation of increased acetylated-histone H3 expression in cells exposed to MSC (Fig. 4A). We hypothesized that a metabolite of MSC, rather than MSC itself, inhibits HDAC activity in human prostate cancer cells. To test our hypothesis, MSP was generated by incubating MSC with LAAO for 2 hours at 37°C and used for HDAC activity assay. HPLC analysis verified that the increase in concentration of MSP closely matched the decrease in concentration of MSC during the incubation (result not shown).

Although MSC had no effect at up to 2.5 mmol/L, HDAC activity was decreased 47% to 52% when nuclear fractions of all four human prostate cancer cell lines were treated with 2.5mmol/L of MSP (Fig. 4B, Upper panel). The inhibition by 2.5 mmol/L of MSP was comparable to that seen with known HDAC inhibitors, TSA (20 μmol/L) and sodium butyrate (2.5 mmol/L). Using the in vitro conditions reported here, we did not observe HDAC inhibition with concentrations lower than 0.025 mmol/L of MSP. This in vitro finding differs from observations under cell culture conditions. We surmise that the disparity may be due to competitive differences between MSP and the fluorogenic histone substrate coupled with developer selectivity and concentration ratio as required by the instructions of the manufacturer. Results similar to those with MSC were observed with selenomethionine, in that selenomethionine had no direct effect on HDAC activity up to 2.5mmol/L, whereas its α-keto acid product, KMSB, exhibited significant HDAC inhibition at 0.25 and 2.5 mmol/L (Fig. 4B, Lower panel). These results corroborate our findings and show the accumulation of acetylated histone H3 when KMSB rather than selenomethionine was introduced directly to cells in culture.

The bacterial A. aeolicus HDAC homologue catalytic domain was built up using the experimentally resolved crystal structure complexed with ligand TSA available in the Protein Data Bank 1C3R (48), and energetically minimized in the internal coordinates space. The model shares most of the residues involved in the zinc atom coordination and in the biological water-mediated catalysis of acetylated lysines with human class I HDACs as well as HDACs from other species (39, 40). To validate the model, TSA was first docked iteratively into the binding pocket, and the most energetically favorable position was compared with the crystal structure (40). TSA docked with an orientation similar to that found experimentally, producing an all-atoms root-mean-square distance value of 0.55 (results not shown).

Next, MSP and KMSB were docked into the HDAC model and the most energetically favorable docking pose obtained was essentially the same for both α-keto acid derivatives (Fig. 2B, left and right). The orientation of the α-keto group and the carboxylate moiety of MSP and KMSB in the pocket closely resembles that seen for the carbonyl group of known HDAC inhibitors, MS-344 and TSA (40) with respect to both zinc atom coordination and the hydrogen bond to the hydroxyl group of Y297. One of the oxygen atoms of the terminal carboxylate group also stabilizes zinc atom coordination in a fashion similar to that reported earlier for sulforaphane- cysteine (49–51). The aliphatic chain of both MSP and KMSB which contains the Se-atom interacts with the phenylalanine group (F141) and this interaction is comparable to that seen in complexes with suberoylanilide hydroxamic acid (40, 51).

Although most former studies on selenium as a nutrient and chemopreventive agent have focused on its role as an essential component of several selenium-containing enzymes (glutathione peroxidases, thioredoxin reductase, iodothyronine 5′-deiodinase) and other selenoproteins (52), recent studies show that small molecular weight organoselenium derivatives may also have intrinsic value. Over the past 10 years, a number of naturally occurring and synthetic organoselenium compounds have been examined for their anticancer properties. It seems that, depending on the molecular form of selenium, selenium compounds possess several modes of action affecting cancer promotion and progression. First, when metabolized as an essential nutrient, the incorporation of selenium into selenoproteins as the 21st coded amino acid, selenocysteine, enables several endogenous selenoenzymes to counteract peroxidative reactions and function as redox catalyzers. Secondly, exogenously administered organoselenium compounds may expose mitotic cells to a pool of novel organoselenium metabolites that have the potential to interact with specific molecular targets necessary for mediating chemopreventive activity. The former chemopreventive mechanisms may be relevant to protecting cells against cancer initiation whereas the latter mechanism may thwart tumor cell progression by maintaining a nonproliferative intracellular environment.

The anticancer mechanism(s) of selenium-enriched diets have focused on two organoselenium nutrients, MSC and selenomethionine, and their potential for being converted to methylselenol via hypothesized β-elimination and γ- elimination reactions, respectively. The anticancer activity of methylselenol is purported to manifest through the oxidation of thiol compounds on proteins and/or enzymes, thereby changing the redox environment of cells (52–56).

Several PLP-containing enzymes are known to catalyze β-elimination reactions, for example, serine deaminase. In addition, cystathionine γ-lyase can catalyze a β-elimination reaction with certain cysteine S-conjugates and with cystine (ref. 57 and references quoted therein). In these two cases, there is no competing aminotransferase reaction. By contrast, several aminotransferases have been shown to catalyze β-elimination reactions with cysteine S-conjugates that contain a good leaving group in the β-position. The β-elimination often competes with the transamination reaction. Thus, as noted in the Introduction, an α-keto acid substrate or PLP is required for the β-lyase reaction to proceed. As also noted, we recently showed that rhGTK can catalyze both transamination and β-lyase reactions with MSC (27). This result suggested that the transaminated product of MSC, in addition to the product of β-elimination, methylselenol, may play critical roles in chemoprevention.

The present study considered the possibility that the α-keto acid products of MSC and selenomethionine would exhibit HDAC-inhibitory properties. We first determined whether human prostate cancer cells metabolize naturally occurring organoselenium amino acids to α-keto acid products and whether these products might increase the levels of histone-H3 acetylation. The α-keto acid products of both MSC and selenomethionine were generated using LAAO and a convenient assay system was developed using reverse phase HPLC coupled with electrochemical detection to identify the seleno α-keto acid products. Both α-keto acids, MSP as a metabolite of MSC and KMSB as a metabolite of selenomethionine, were verified as the products of the LAAO reaction using MS analysis and further analyzed for HDAC-inhibitory properties using a fluorescent-based assay for measuring class I and II HDAC activity. Accordingly, each α-keto acid product inhibited HDAC activity in a dose-dependent manner using a HeLa cell nuclear extract provided with the commercial kit (data not presented).

Both androgen-responsive (LNCaP) and androgen- independent (LNCaP C4-2, PC-3, and DU-145) cell lines were used to test our hypotheses that α-keto acid products of MSC and selenomethionine function as HDAC inhibitors in a whole cell environment. First, we determined whether prostate cancer cells possess GTK activity and have the ability to transaminate MSC. Using the standard aminotransferase substrates for GTK, phenylalanine, and KMB, all four cell lines exhibited GTK activity. Curiously, although methionine functions as an excellent transamination substrate for rhGTK, selenomethionine is <0.3% as reactive (27). This was further corroborated when whole cell lysates from each of the prostate cancer cell lines were incubated with MSC and selenomethionine, and aliquots removed after 2 hours and analyzed by HPLC to detect their corresponding α-keto acid metabolites. Only MSP could be detected, whereas selenomethionine was not metabolized to KMSB even after 24 hours of incubation. In order not to negate the long-term effects of selenomethionine when administered orally to animals, we are investigating the possibility that dietary forms of selenomethionine (from general protein sources or its free form) may undergo hepatic trans-selenation reactions to form methylselenocysteine and thus contribute to the seleno α-keto acid product via this pathway (52).

Treatment of LNCaP, LNCaP C4-2, PC-3, and Du-145 prostate cancer cells with MSC led to time-dependent and dose-dependent changes in histone acetylation status. Treatment of cells with MSC resulted in an increased accumulation of acetylated histone H3 in LNCaP cells as early as 5 hours and in the androgen-independent cell lines within 24 hours postincubation. Cells treated with selenomethionine did not show acetylated histone H3 accumulation even after 24 hours of incubation. When cells were directly treated with the seleno α-keto acid products, MSP and KMSB, acetylated histone H3 accumulated as early as 5 hours posttreatment. We are currently investigating the possibility that longer incubation times (>48 hours) may be required to generate sufficient concentrations of KMSB necessary to elicit HDAC inhibition. Preliminary enzyme kinetic data suggest that the Ki of KMSB for HDAC8 inhibition may be ∼30 μmol/L (58). These findings are consistent with our previous finding that MSC is >100 times more effective as an aminotransferase substrate for rhGTK compared with selenomethionine despite the difference of only one methylene group (27).

By contrast, treatment of cells directly with the α-keto acid products of MSC and selenomethionine result in increased accumulation of acetylated histone H3 in as early as 5 hours after incubation. Selenomethionine is not directly metabolized by human prostate cancer cells to KMSB. KMSB can be detected in liver homogenates incubated with selenomethionine, suggesting that liver tissue may contain other aminotransferases or an LAAO as well as cystathionine γ-lyase, (although not detected in prostate cancer cells), which may catalyze deamination or γ-elimination products, respectively, from selenomethionine. A question of residence time, stability, and extent of plasma circulation will be important factors to consider in subsequent studies, including potential follow-up to the recently halted SELECT study.

At present, we cannot conclude from our in vitro studies whether sufficient plasma levels of the α-keto acid products of MSC and selenomethionine can be achieved in preclinical models of prostate cancer over time with daily consumption or whether unequivocal HDAC inhibition occurs in vivo when dietary sources of selenium in the range of 2 to 6 ppm are consumed. This caveat withstanding, most tissues, with the exception of erythrocytes, contain aminotransferases potentially capable of metabolizing MSC or even selenomethionine (27); however, pharmacokinetic studies need to be conducted to determine the bioavailability and metabolic fates of these compounds using our HPLC electrochemical method. As mentioned earlier, preliminary studies using an in vitro assay of HDAC activity suggest that the IC₅₀ may be in the range of 30 to 50 μmol/L, a level that may be achievable following the administration or consumption of the preformed α-keto acid products. These effects remain to be established in a preclinical model of prostate cancer.

In conclusion, we have identified two metabolites of naturally occurring organoselenium derivatives as novel HDAC inhibitors. We first developed an assay system for these α-keto acid metabolites using HPLC with electrochemical detection and identified one of these metabolites (MSP) within prostate cancer cells. The findings for MSP and KMSB with respect to HDAC inhibition and induction of acetylated histone H3 in prostate cancer cells extend our current understanding of how organoselenium may decrease the progression of human prostate cancer. By altering posttranslational modifications of histones, HDAC inhibitors facilitate the access of transcription factors to DNA, and enable the unsilencing of tumor suppressor genes in cancer cells (59). This paradigm has been viewed from the perspective of developing novel drugs for cancer therapy, but is now expanding to include dietary chemopreventive agents (49, 50, 58, 60). The next goal will be to define protein targets of seleno α-keto acids and the underlying mechanisms, which might help to refocus efforts in the wake of the recently halted SELECT study.

No potential conflicts of interest were disclosed.

We thank Heather Twerdahl and Lawson Kurtz for technical assistance.