Synthesis and anticancer activity of new flavonoid analogs and inconsistencies in assays related to proliferation and viability measurements

Flavonoids have been studied intensely for their ability to act as anti-carcinogenic, anti-inflammatory, anti-viral and anti-aging agents and are often marketed as supplements related to their anti-inflammatory activity. Previous studies have primarily focused on the effects of polar natural flavonoids. We examined the activity of novel hydrophobic and lipophilic flavonols against human DU-145 and PC-3 prostate cancer cell lines. All flavonol analogs were more active than the naturally occurring flavonols quercetin, kaempferol, kaempferide and galangin. The most potent analogs were 6.5-fold more active against DU-145 and PC-3 cells than quercetin and fell within the biologically relevant concentration range (low micromolar). We also evaluated the potential toxic effects of flavonol analogs on normal cells, an assessment that has frequently been ignored when studying the anticancer effects of flavonoids. During these analyses, we discovered that various metabolic and DNA staining assays were unreliable methods for assessing cell viability of flavonoids. Flavonoids reduce colorimetric dyes such as MTT and Alamar Blue in the absence of cells. We showed that flavonol-treated prostate cancer cells were stained less intensely with crystal violet than untreated cells at non-toxic concentrations. The trypan blue exclusion assay was selected as a reliable alternative for measuring cell viability.


Introduction
Numerous papers exist regarding the effects of flavonoids aiding in the prevention of cancer growth through their ability to act as anti-oxidants (1), enzyme inhibitors (2) and growth regulators (3). However, many of these papers limit their conclusions to findings on cancer cells and fail to test the impact flavonoids have on normal cells (4)(5)(6)(7). Studies, dating back to 1977, report a connection between high flavonoid concentrations and potential mutagenic tendencies (8)(9)(10)(11)(12)(13). Skibola and Smith (8), in 2000, addressed the potential danger of consuming high concentrations of flavonoids through supplements, which are not regulated by the FDA.
The flavonol, quercetin, is a common supplement promoted as having the ability to aid in cancer prevention, treat chronic infections of the prostate and other common ailments (14)(15)(16). Humans typically consume between 4 to 68 mg of flavonols daily through the consumption of fruits and vegetables (8). Flavonoid supplements, which range from 500 to 1000 mg per tablet, can provide approximately 15 times more of the flavonol than can be obtained through dietary consumption. Therefore, it is important to determine flavonoid toxicity in normal cells in order to establish if flavonoid supplements are truly safe for human consumption. Additionally, one must consider the rate of flavonoid absorption and bioavailability when designing a study to examine flavonoid activity in vitro (14,(17)(18)(19)(20). Flavonoids with concentrations above the absorption and bioavailability limit are not relevant to real world applications, since they will not be absorbed by the body.
Flavonoids have previously been studied as potential therapeutic agents for breast (21,22), prostate (23), lung (24,25), colon (26) and skin (27) cancers. To be a good potential therapeutic agent, flavonoids must be able to reduce cell viability in the cancerous cells, while having a minimum effect on the normal cells. In prostate cancer, polar natural flavonols (fisetin, galangin, kaempferol, morin, myricetin and quercetin) have commonly been found to exhibit these characteristics (3,(28)(29)(30). However, there is limited information about the effects of hydrophobic and lipophilic flavonols on prostate cancer. The more hydrophobic (alkoxyl, geranyl > dimethylallyl > halogen > monolignol > methoxy > hydroxyl > glycosyl) and lipophilic flavonols (I > Br > Cl > F) are the most potent inhibitors of P-glycoprotein (P-gp), which is an important protein involved in drug sensitivity and resistance (31). Halogenated flavonols could also interact with Lewis bases, such as amines or alcohols of amino acids, to potentially form non-covalent halogen bonds within the cancer cells (32). Based on this limited information, a series of more hydrophobic and lipophilic analogs were synthesized with the assumption that they would decrease cell viability of DU-145 and PC-3 prostate cancer more effectively than the polar natural flavonols. The effects of flavonols on the viability of normal human infant foreskin fibroblasts (HIFF) were also studied.
Flavonoid cell viability has most commonly been measured through colorimetric assays. However, in 2002, Bruggisser et al (33) found MTT to be an unreliable technique for studying flavonols due to the ability of flavonols to reduce MTT in the absence of cells. Since then, at least 1019 flavonoid articles have been published using the MTT assay. (A SciFinder search was conducted on 01/22/14 and consisted of keyword: flavonoids; refine: MTT; sort by: publication year) A recent review of the literature indicated that Alamar Blue, crystal violet and trypan blue had also been used to determine flavonoid cytotoxicity (Fig. 1). Thus, we evaluated the reliability of these lesser utilized methods in the presence of hydrophilic, hydrophobic and lipophilic flavonols to find a more accurate method of measuring flavonoid cytotoxicity.

Materials and methods
Chemistry reagents. All chemicals and solvents were purchased from Sigma-Aldrich (St. Louis, MO, USA). Spectra were obtained on a Perkin-Elmer Paragon 1000 FT-IR spectrometer. Proton and carbon NMR spectra were recorded on a Varian Mercury Vx 300 or 500 MHz spectrometer with (CD 3 ) 2 CO and CD 3 OD as the solvents. High resolution mass spectral (HRMS) data were obtained on an Applied Biosystems/MDS SCIEX 4800 Plus MALDI TOF/TOF mass spectrometer. Melting points were determined on a Thomas Hoover Uni-melt and are uncorrected. The purity of the compounds was determined to be >95% by elemental analysis (Galbraith Laboratories, Inc., Knoxville, TN, USA).
General procedure for the synthesis of the flavonol analogs. Synthesis of the flavonol methyl ethers. Methoxyphloroacetophenone (50 mg, 2.53 mmol) was placed into a round bottom flask, dissolved in 1,2-dichloroethane (5 ml) and 4 ml were distilled to remove water. The carboxylic acid (3.54 mmol) and the acid chloride (8.33 mmol) were then added and dissolved in triethylamine (5 ml) and dimethylformamide (DMF) (15 ml). The solution was stirred at room temperature under argon for 3 h before being heated to reflux for an additional 12 h. The triethylamine and DMF were then removed under vacuum. The residue was dissolved in 3.8 ml of ethanol and heated to reflux. A solution of KOH (1.0 g) in H 2 O (1.4 ml) was then added and the mixture was refluxed for an additional 60 min. The solvents were removed under vacuum. The solid was dissolved in water (30 ml) and the crude product was precipitated out of solution by the addition of 10% HCl until the solution reached a pH of 5.0. The crude precipitate was extracted into chloroform (5 ml) and rinsed with sodium bicarbonate (10 ml x 2). Since the desired product was only partially soluble in chloroform, these last two steps must be repeated several times (6-10 x: until the chloroform extractions are no longer yellow). The chloroform was removed under vacuum. The crude product was then dissolved in a minimal amount of acetone (~0.5 ml per 10 mg of flavonol). The solution becomes cloudy upon the addition of hexanes (10 ml) and was filtered through a pipet packed with glass wool. The filtrate contains the desired yellow product. The filtrate was evaporated under vacuum and set aside. Since some of the product remained in the precipitate, the precipitate on the glass wool in the pipet was dissolved in acetone, the acetone was evaporated, and trituration with acetone/hexanes was repeated (2-4 x: until the filtrate is no longer yellow). The NMR spectrum confirmed that the entire product was removed from the crude precipitate.
Deprotection: The methyl ether flavonol (20 mg, 0.055 mmol) was placed into a round bottom flask and dissolved in acetic acid (16 ml). HBr (48% aqueous solution, 2 ml) was added via a syringe. The solution was heated to reflux for 24 h. Water (150 ml) was then added, resulting in a cloudy solution. The precipitate was isolated by filtration. This crude product was dissolved in chloroform (10 ml) and rinsed with sodium bicarbonate (10 ml x 3). The chloroform was evaporated under vacuum. The crude product was dissolved in a minimal amount of acetone (0.5 ml) and an excess of hexanes (10 ml) was added. The filtrate was collected and the process was repeated several times. The filtrate was dried under vacuum to give a yellow precipitate.
2-(3-Benzoylphenyl)-5,7-dihydroxy-3-methoxy-4Hchromen-4-one (Fig. 2   2-(3-Benzoylphenyl)-3,5,7-trihydroxy-4H-chromen-4-one (3'-BPK) (Fig. 2    Treatment and measurement of cell viability. Stock solutions of the natural product and synthetic flavonol analogs were prepared by dissolving them in DMSO. Preliminary doseresponse studies were conducted in the prostate cancer and HIFF cell lines with each flavonol to find the minimum and maximum effective dose range and an approximate EC 50 . The actual EC 50 for each flavonol was calculated from a minimum of 6 concentrations selected within the linear portion of initial dose range curves and bracketing the estimated EC 50 . Cells were seeded into 24-well plates 24 h prior to the experiment at the following cell densities: DU-145 (6,000 cells/well), PC-3 (6,500 cells/well) and HIFF (6,000 cells/well). The stock solutions of the flavonols were added to each of the wells, except the control, to give the desired concentration of the flavonols and a final concentration of 0.2% DMSO and the plates were incubated for 72 h. DMSO did not alter the growth or viability of DU-145, PC-3 or HIFF (data not shown). Assays were run in quadruplicate. The reliability of three cell viability assays was tested with the following reagents: i) Alamar Blue: After removal of the culture medium, cells were washed with PBS and incubated with Alamar Blue for 6 h. Fluorescence was measured with an excitation at 530 nm and an emission at 590 nm. ii) Crystal violet: After removal of the culture medium, the cells were fixed with 200 µl of 4% formaldehyde at 4˚C for 30 min, washed with a 3:1 mixture of MeOH: AcOH, rinsed with 80% MeOH, and stained with 0.1% crystal violet. After 1 h at room temperature, the wells were gently rinsed with distilled water, dried, dissolved in 10% AcOH and the absorbance was read at 570 nm. iii) Trypan blue exclusion: The cells were trypsinized, washed with calcium free phosphate buffered saline (PBS), centrifuged and resuspended in 500 µl of PBS. Cell viability was accurately determined using the semi-automated Vi-Cell XR Cell Analyzer (Beckman Coulter).
Statistical analysis. The EC 50 values of the flavonols in each cell line were analyzed by one-way analysis of variance (ANOVA). Comparisons of the EC 50 values of the flavonols between cell lines were evaluated by two-way ANOVA. Differences were considered significant at P<0.05.

Chemistry.
A series of flavonol analogs were synthesized from the starting material, ω-methoxyphloroacetophenone 1 (Fig. 2) (23,35,36). Initially, these compounds were synthesized through a modified multistep procedure by Tanaka et al (37), in which the ω-methoxyphloroacetophenone 1 was esterified by the acid chloride 3 to produce a triester intermediate, which was subsequently cyclized by refluxing in pyridine containing K 2 CO 3 to give the flavonol methyl ether 4. This sequence resulted in moderate net yields of the flavonol methyl ethers. A major reason for the modest yields was the poor recovery of the products from the chromatographic purifications at each step. Therefore, we explored an alternative approach, the 'one pot' procedure described by Ichikawa et al (38) followed by ester hydrolysis, to synthesizing the flavonol methyl ether 4. Ichikawa et al used Et 3 N as both the solvent and the base. This procedure resulted in slightly higher yields (26-35%) of the flavonol methyl ether 4 than the previous process. During these reactions we noted the low solubility of the ω-methoxyphloroacetophenone in Et 3 N. Based on these observations, the yields of compound 4 were increased through a series of optimization reactions (solvent, temperature and reaction time). These studies led to the use of Et 3 N as the base and DMF as the solvent to give a homogenous solution, which significantly increased the yields of the cyclized products. With these optimal conditions, good yields were achieved in the synthesis of the hydrophobic aromatic and halogenated flavonol ethers 4a-d [3'-benzoyl (55%) and The remaining two analogs (3'-and 4'-phenyl) were synthesized through a Suzuki-Miyaura palladium catalyzed cross-coupling reaction using a procedure developed by Molander et al (34) (Fig. 3). The 3'-and 4'-iodo flavonol ethers 4c,d readily coupled with potassium phenyl trifluoroborate using Pd(OAc) 2 and potassium carbonate in methanol to give high yields of both 3'-phenyl 7c (68%) and the 4'-phenyl 7d (67%) flavonol ethers. These compounds were readily demethylated to the 3'-phenyl 8c (93%) and 4'-phenyl 8d (98%) flavonols using the previously determined reaction conditions.
Biological method validation of cell viability assays. The reliability of various colorimetric and cell counting assays in the presence of flavonols was determined by comparing the metabolism-dependent Alamar Blue assay, the DNA-binding crystal violet assay and the dye exclusion trypan blue method. We found that flavonols reduced Alamar Blue in the absence of cells and overestimated the number of viable prostate cancer cells (Fig. 4). Flavonols also interfered with the crystal violet assay by staining untreated cells more intensely than flavonol treated cells. For many flavonol treated cells, the maximum achievable cell viability was only 50% compared to the control, even at non-cytotoxic concentrations. For some flavonols, such as kaempferide, the maximum cell viability did not ever reach 50%.
The trypan blue cell counting assay proved to be the most reliable method for measuring cell viability in the presence of flavonols with a minimal difference (≤10%) between the total number of cells counted and the number of cells stained with trypan blue. These results indicated that flavonols did not interfere with the cell counting readings and provided a model by which to compare the reliability of other assays such as Alamar Blue and crystal violet. A comparison between trypan blue and Alamar Blue cell viability curves showed a shift in the Alamar Blue curve to the right, which is in agreement with previous studies that phenolic compounds resulted in an overestimation of cell viability (Fig. 4). Quercetin had an EC 50 value of 55.1 µM in Alamar Blue compared to 39.1 µM in trypan blue. Kaempferide had an EC 50 value of 30.1 µM in Alamar Blue compared to 19.8 µM in trypan blue. We also found, for some compounds, that the reproducibility of Alamar Blue was poor (Fig. 5). This may have resulted from inconsistent flavonol reduction of the metabolic compound between analyses with fluctuating variables, such as the amount of flavonols remaining in the well during incubation and cell numbers between analyses. The major finding using the crystal violet assays was that flavonols significantly  masked the effects on cell viability readings at lower concentrations (Fig. 4). However, at higher flavonol concentrations this interference appeared to be minimal and comparable to the trypan blue assay.

Effects of flavonols on the in vitro viability of DU-145 and PC-3 prostate cancer cells.
Flavonols with a carbonyl at C4, a double bond between C2 and C3, and a hydroxyl group at position 3 of ring C, were selected for this study to determine cytotoxicity towards androgen-independent DU-145 and PC-3 human prostate cancer cells in comparison to toxicity towards HIFF normal cells as determined using the trypan blue exclusion assay. All of the flavonols tested decreased the number of viable DU-145 and PC-3 cells. In DU-145 cells, the more hydrophobic and lipophilic flavonol analogs (3'-and 4'-phenyl and 4'-iodo flavonols) were the most potent (Table I). The other hydrophobic and lipophilic flavonols (3'-iodo, 3'-and 4'-benzoyl and 4'-methyl flavonols) as well as the natural product, kaempferide, were slightly less efficient at decreasing cell viability. The natural product, galangin, moderately decreased viability, while quercetin and kaempferol had the least effect.
A comparison of the differences in the EC 50 values of the flavonols between the two prostate cancer cell lines showed that only three compounds, quercetin, galangin and 4'-benzoyl flavonol, exhibited differential cytotoxicity (Fig. 6). All three    (Table II). The more hydrophobic and lipophilic analogs (3'-and 4'-phenyl, 3'-and 4'-iodo, 3'-benzoyl and 4'-methyl flavonols) had the lowest EC 100 values. The natural products, kaempferide and galangin, were slightly less efficient at achieving their maximum effect. Quercetin and kaempferol were the least effective. A comparison of the differences in EC 100 values of flavonols between these two cell lines showed that only two compounds, kaempferol and 4'-benzoyl, were statistically different (data not shown). Both compounds were more effective against PC-3 than DU-145 cells (P<0.05). From these results, we concluded that only a minimum concentration (<25 µM) of the more hydrophobic and lipophilic flavonol analogs (3'-and 4'-phenyl, 3'-and 4'-iodo, 3'-benzoyl and 4'-methyl flavonols) was needed to achieve a maximum effect on both the DU-145 and PC-3 cell lines.
A comparison of the difference between the EC 100 and EC 50 value of the flavonols in DU-145 showed that only a small increase in concentration (10 µM or less) was needed to achieve maximal effect for the hydrophobic and lipophilic analogs (4'-phenyl, 3'-benzoyl, 4'-methyl, 3'-and 4'-iodo flavonols) (Table III). The natural products, galangin and kaempferide, as well as the flavonol analog, 4'-methyl flavonol, had a slightly larger range (<25 µM) with 3'-phenyl flavonol overlapping the two categories. Quercetin and kaempferol were the least effective and the range between the EC 100 and EC 50 values was at least a 36 µM.
In PC-3 cells, a 10 µM or less difference between the EC 100 and EC 50 values was found in the more hydrophobic and lipophilic flavonol analogs (4'-phenyl, 3'-benzoyl, 3'-phenyl, 3'-and 4'-iodo flavonols) ( Table III). The less hydrophobic natural products, kaempferol, galangin and kaempferide, required a slightly larger range (<25 µM) to achieve a maximal effect with the 4'-methyl and 4'-benzoyl flavonol overlapping the two categories. Quercetin was the least effective and the range between the EC 100 and EC 50 values was 44 µM.
A comparison of the differences in the EC 100 and EC 50 values of flavonols between these two cell lines showed that only two compounds, kaempferol and 4'-benzoyl flavonol, were statistically different (data not shown). These results indicate that hydrophobic and lipophilic flavonol analogs tend to reach their maximal effect at lower concentrations than natural flavonols and that this maximal effect is achieved over a more narrow effective concentration range (with 4'-benzoyl flavonol being the exception in DU-145). Narrow ranges between the EC 50 and EC 100 values are beneficial because they reduce the likelihood that the agents will be toxic to normal cells.  Effects of flavonols on the in vitro viability of HIFF normal cells. HIFF cells were used to determine the toxicity of flavonols on normal cells. Flavonoid research has commonly been conducted at concentrations of 50 µM or greater (5,46). Based on this information, we first examined the effects of flavonols on HIFF cells at 50 µM concentrations. At this concentration, we found that all flavonols tested caused a decrease in cell viability (Table IVA). Quercetin was the least inhibitory and all of the synthesized derivatives exhibited major inhibitory activity at the 50 µM concentration. HIFF cell viability was also examined at the average flavonol EC 50 concentrations (for DU-145 and PC-3) for individual compounds. At this concentration prostate cancer cells would have a cell viability of 50%. Kaempferol, 4'-benzoyl and 3'-iodo flavonols had no affect on the cell viability of HIFF cells (Table IVB). The natural products, quercetin and galangin, and the flavonol analog, 3'-benzoyl flavonol, caused a slight decrease in the viability of HIFF cells. A moderate effect was observed for kaempferide and 4'-phenyl flavonol. Three flavonols, 4'-methyl, 4'-iodo and 3'-phenyl flavonol exhibited the strongest decrease in cell viability. The activity of flavonols towards HIFF cells did not correlate with the activity of flavonols towards the prostate tumor cells. A comparison of the flavonols, which minimally decreased the viability of HIFF cells and efficiently decreased the viability in prostate cancer cells at both the EC 50 and EC 100 values, lead us to conclude that 3'-iodo and 3'-benzoyl flavonols are likely to be the most promising therapeutic agents.

Discussion
For decades, flavonoids have been extensively studied for their ability to act as anti-oxidants, enzyme inhibitors and growth regulators. More recently, flavonols have become available as supplements. In preparing to examine the effects of new flavonol analogs on cell viability of human prostate cancer cells, we found that many previous reports failed to use reliable methods to examine this effect and in addition, a number of studies examined the effects of non-biologically relevant concentrations of flavonoids on cell viability. Cell viability and proliferation (47)(48)(49)(50)    and MTS (57), a second generation tetrazolium dye, in the absence of cells and CTG (22) overestimates the number of viable cells for some polyphenolic compounds. In this study, we found that flavonoids also reduce Alamar Blue in the absence of cells and overestimate the number of viable cells. For NAD(P)H-dependent assays this interference is likely due to the ability of flavonoids to mimic the reducing agent, NAD(P)H. Flavonoids are known anti-oxidants, capable of preventing oxidative damage by being oxidized themselves (58)(59)(60). Several anti-oxidants (flavonoids, ascorbic acid, vitamin E, ubiquinone, hydroquinone and N-acetylcysteine) also reduce MTT in the absence of cells suggesting that MTT and other metabolic-dependent agents are not reliable assays for analyzing agents that have anti-oxidant activity (33,(61)(62)(63)(64). For ATP-dependent assays, flavonoids likely mimic ATP by binding to the ATP binding site (2,29,31). Several flavonols, such as quercetin, kaempferol and galangin are known to exhibit this property.
Flavonoids also interfere with the staining of DNA (Hoechst 33342, propidium iodide and crystal violet) and proteins (rhodamine 123). For example, the flavonol, quercetin, previously was found to inhibit the binding of Hoechst 33342 to DNA and to activate the binding of rhodamine 123 to proteins (31,65). The anthocyanin, cyanidin-3-rutinoside, prevented the binding of propidium iodide to DNA (66). These dyes have commonly been used in flow cytometry studies. We found that crystal violet stained the DNA of untreated cells more intensely than flavonol treated cells. In several cases, the maximum achievable cell viability was only 50% of the control at nontoxic concentrations. The interference of the flavonoids with the ability of crystal violet to bind the DNA might have resulted from intercalation of the flavonols into DNA, since flavonols such as quercetin and kaempferol are known to exhibit this property (67). Rhodamine 123 measures mitochondrial membrane potential, which is regulated by ATP (68). Therefore, if quercetin binds to the ATP binding site it would disrupt the membrane potential and would interfere with the accumulation of rhodamine 123 inside the membrane. Additionally, quercetin enhances the rate of rhodamine 123 transport into P-glycoproteins (69).
In order for these assays to be reliable, excess flavonoids must be removed before the addition of the metabolic agent. Previous investigators have suggested removing the culture medium before the addition of the metabolic agent (33). However, flavonoids are minimally soluble in biological buffers. Therefore, this method is inefficient at removing all of the flavonoids in solution and in addition, does not account for the flavonoids located within the cell that can also interfere with the colorimetric agent. Liu et al (70) previously found in B12 cells, which are phenotypically similar to neural precursor cells, that genistein reduced 60% more MTT than the control during a 3 h incubation. Therefore, the impact that flavonoids have on metabolic agents may vary depending on the amount and type of flavonoid used.
Based on these results, we concluded that, though popular, metabolic and DNA staining assays are not a reliable method of measuring cell viability and proliferation in the presence of flavonoids. The trypan blue cell counting assay provides an alternative method to these assays, but has its own set of limitations (71)(72)(73). Some color variations may be undetectable to the human eye and debris can also be stained and counted as a nonviable cell. For accurate results, early phase cells (≤72 h) must be used to prevent an overestimation of the number of viable cells and cells must be counted within 5 min of the addition of the dye. An automatic cell viability analyzer such as the Vi-Cell XR (Beckman Coulter Inc.) addresses most of these limitations by more accurately detecting different sizes and shades of stained cells, allowing rapid analysis (<3 min) and providing more consistent results than the manual method (74). Additionally, the difference between the total number of cells and the number of nonviable cells was ≤10%, implying that flavonols did not interfere with the cell counting readings. Our results correlate well with previously published trypan blue results. For example, Vijayababu et al (75) found in PC-3 cells that quercetin had an EC 50 value of around 29 µM compared to our 32.9 µM.
Polar natural flavonols (quercetin, kaempferol, kaempferide and galangin) have commonly been the focus of flavonoid cancer research. However, their in vitro anticancer effects have not translated well to their effects in vivo, probably, because absorption and bioavailability of dietary flavonoids (4-68 mg) are limited to the low micromolar range (76)(77)(78) and previously, their effective and cytotoxic dosages have been well above this concentration limit. For example, the flavonol supplement quercetin has an EC 50 value of 39.1 µM in DU-145 and 32.9 µM in PC-3 cells. We found that increasing the hydrophobicity and lipophilicity of a flavonol significantly lowers the effective concentration of several analogs (4'-iodo, 3'-phenyl and 4'-phenyl flavonol) into the desired low micromolar range thus enhancing their therapeutic potential. However, information about the absorption and bioavailability of these novel flavonols are needed to provide a more complete picture of their ultimate anticancer potential in vivo.