Measuring the Chemical and Cytotoxic Variability of Commercially Available Kava (Piper methysticum G. Forster)

Formerly used world-wide as a popular botanical medicine to reduce anxiety, reports of hepatotoxicity linked to consuming kava extracts in the late 1990s have resulted in global restrictions on kava use and have hindered kava-related research. Despite its presence on the United States Food and Drug Administration consumer advisory list for the past decade, export data from kava producing countries implies that US kava imports, which are not publicly reported, are both increasing and of a fairly high volume. We have measured the variability in extract chemical composition and cytotoxicity towards human lung adenocarcinoma A549 cancer cells of 25 commercially available kava products. Results reveal a high level of variation in chemical content and cytotoxicity of currently available kava products. As public interest and use of kava products continues to increase in the United States, efforts to characterize products and expedite research of this potentially useful botanical medicine are necessary.


Introduction
Kava (Piper methysticum G. Forster) is the name of a plant and drink that is prepared traditionally by macerating its roots in cool water or coconut water [1]. It has been used for many centuries in the South Pacific and Hawaii for social ceremonies, relaxation, medicine, and a multitude of other purposes [1]. More recently, standardized kava extracts, containing 30% active constituents, have been used globally as an anxiolytic [2,3]. Additionally, a tight inverse correlation between high rates of kava consumption and low incidences of cancer for populations in the South Pacific has been reported [4]. Subsequent studies have shown that kava displays cancer preventive properties [5][6][7][8].
A 2002 the Kava Act passed in Vanuatu established four classes of kava cultivars: noble, which have a long history of safe use as traditional drink; medicinal, which have long been used by traditional herbalists in the South Pacific and are banned as export commodities; 'Tu dei', which have a very strong effect that lasts two days; and 'Wichmanni' or wild varieties [9,17,18]. Cultivars from the noble class are typically used to prepare kava extract as they have the optimal therapeutic chemotype. Cultivars belonging to other classes have been reported to have overpowering and unpredictable effects causing symptoms such as nausea and headaches [19,20]. Kava's active constituents are primarily located in its roots; other plant parts such as stems and leaves should not be used in extract preparations [1]. Traditional kava is prepared using a 100% aqueous solvent, which results in a drink containing an average of 0.3-20% kavalactone content [1]. Commercial manufacturers use up to 100% ethanol or acetone in the extraction process resulting in up to 70% kavalactone content in the final product [1,20]. Studies have shown a difference between traditionally prepared extracts and those prepared with ethanol both in cytotoxicity and chemical composition [21][22][23][24][25]. In addition to those described above, other potential sources of variation in kava products include, contamination of raw kava materials, impurities, post-harvest handling and storage procedures (drying, whole vs. ground material, humidity, temperature), age of harvested kava plants, mixture and quality of cultivar(s) used [17,[26][27][28] Kava is distributed in variable forms, including dry powder, freeze-dried, liquid tincture, and capsule, making it difficult to know exactly which cultivar(s), plant part(s), extraction solvent(s), and other factors were used in the preparation [26].
Due to reports linking modern kava consumption to individual cases of hepatotoxicity, kava was banned in the European Union and Canada in 2003, voluntarily recalled in Australia in 2003, and included on the United States Food and Drug Administration (US FDA) consumer advisory list in March of 2002 [14,[29][30][31][32]. These bans and advisories have hindered research on kava as an alternative anti-anxiety and cancer preventive medicine [17]. Despite its presence on the US FDA consumer advisory list for the past decade, the extrapolation of export data from the kava producing nations Fiji, The Republic of Vanuatu, and Tonga to the US (Table 1) indicates that kava imports to the US, which are not publicly reported, are presumably both increasing and of a fairly high volume [33][34][35][36].

Results and Discussion
Metabolic fingerprinting experiments measured three aspects of chemical variation: reproducibility of replicate extractions of individual products; differences between using 100% water or 95% ethanol as the extraction solvent; and the overall variation among the set of kava products tested. Similarly to previous quantitative studies of compounds from kava, we observed a high level of reproducibility of replicate extractions of material from individual kava sources [45]. Principal component analysis showed that replicate extractions from the same kava source are tightly clustered ( Figure S1). Moreover, there were small standard errors (average standard error 12.6%) from the absolute quantification measurements of K, DHK, M, DHM, FLKA, and FLKB (Tables  S2 and S3). These results provide evidence that there is consistency in the material contained within a single batch of kava from any given source.
Extract chemical composition was strongly influenced by extraction solvent. Metabolic fingerprints from aqueous and ethanolic extracts plotted in principal component space formed two distinct groups driven by extraction solvent where the use of either 100% water or 95% ethanol was responsible for 71.1% of the variation among all samples explained by PC1 ( Figure S2). The detected ion m/z, retention time pairs that contribute the most to the loadings for PC1 were 315.1132 m/z, 9.1823 min and 285.1021 m/z, 9.4699 min, which correspond to the masses and retention times of FLKA and FLKB, respectively. Compound quantification showed that extracts prepared with 95% ethanol resulted in higher yields and greater consistency among replicates, compared with extracts prepared with 100% water. This result is similar to previous studies that found water produced kava extracts with decreased compound concentrations compared to extracts prepared with ethanol [22,23]. Specifically, K, DHK, M, and DHM concentrations were 1.5-5x higher in samples extracted with 95% ethanol than in those extracted with 100% water. The concentrations of FLKA & FLKB were up to fifty times higher in samples extracted by 95% ethanol than in those extracted with 100% water although a significant number of the water extracts contained concentrations of FLKA or FLKB that were below detectable limits ( Figure 1). Extracts prepared with 95% ethanol consistently contained greater quantities of FLKA and FLKB than corresponding water extracts, and were highly variable across kava products (ranging from undetectable concentrations up to 14.7 ppm; Figure 1). While extraction solvent was the most influential variable affecting the observed chemical composition, significant variation in the concentrations of K, DHK, M, and DHM for identically prepared extracts was observed from different source materials. This variation was even more dramatic in regard to the concentration of FLKA and FLKB. The variation in chemical composition was further reflected by the differences in cytotoxicity observed for each commercial kava product extract.
Cytotoxicity assays against human lung adenocarcinoma A549 cancer cell line with aqueous extracts from all 25 commercial kavas showed no toxicity at any concentration measured up to 500 mg/mL. This result is similar to previous studies indicating that aqueous extracts have low to no cytotoxic effect [20]. In contrast, identically prepared ethanol extracts from different commercial sources varied greatly in their relative cytotoxicity at all concentrations measured 37.5, 75 (shown in Figure 2 top), and 150 mg/mL. Ethanol extracts prepared from commercial kava sources K, M, N, O, W, Y, DD, and EE exhibited very low cell toxicity at all concentrations, indicated by a relative cell viability level of greater than 90%. In contrast, ethanolic extracts from G, H, P, Q, R, S, V, Z, and BB, displayed the highest levels of toxicity, with a relative cell viability level of less than 30%. Cytotoxicity levels at these three discreet extract concentrations varied over a wide range similarly to the variation observed in extract chemical composition, especially in regard to FLKA and FLKB.
We observed a moderate correlation between the concentrations of FLKA & FLKB (log 2 normalized) and the relative cytotoxicity across the sampled kava products. High concentrations of the flavokawains generally mirrored lower relative cell viability ( Figure 2). Some samples, however, deviated from this correlation, specifically N and BB, which have medium concentrations of FLKA and FLKB and display low and high cytotoxcity, respectively. Scatterplots of these data revealed that concentrations of FLKA and FLKB correlated similarly to cytotoxicity level with R 2 values equal to 0.68 and 0.69, respectively for extracts prepared at 75 mg/mL and 0.78 and 0.77, respectively for extracts prepared at 150 mg/mL (Figures S3 and S4). Based on these correlation values, clearly, FLKA and FLKB are likely the major but not the solely compounds responsible for the extract toxicity. However, additional reports of flavokawain toxicity, including measured IC 50 levels for FLKA and FLKB of 1361.1 and 6.660.1 ppm, respectively against Hepa 1c1c7 liver cells [24], and 57% growth inhibition of bladder T24 tumor cells by FLKA [8] warrant further exploration of the link between flavokawains and kava's cytotoxicity [21,22].
We also generated a non-polar flavokawain enriched kava fraction to determine the IC 50 values of 48-hour toxicities against hepatocytes from mouse, rat, and monkey. We measured IC 50 values for this FLK rich fraction of 5769, 4564 and 4966 mg/ mL, for mouse, rat, and monkey hepatocytes, respectively. In contrast polar and medium polarity fractions and whole traditionally prepared kava had non-detectable IC 50 values greater than 400 mg/mL in all three cell types.
For each compound, K, DHK, M, and DHM there was no obvious association between concentration and relative cell viability, although extracts with higher overall concentrations of all six compounds resulted in lower relative cell viability. This trend suggests that K, DHK, M, and DHM are less likely to be involved in the specific mechanism(s) of cytotoxicity.
Additional reports of cytotoxic compounds found in kava describe one additional flavokawain, distinct from FLKA and FLKB with the following chemical formula and exact mass C 17 H 16 O 5, 300.0998 (FLK C) [35] and three alkaloids found in kava leaves with the following chemical formulae and exact masses: C 14 H 17 NO 2 , 231.1259 (awaine); C 16 H 17 NO4, 287.1157 (pipermethystine); C 16 H 17 NO 5 , 303.1106 (3a, 4a-epoxy-5b-pipermethystine) [47]. These alkaloids may have been present in commercial kava products produced by European companies leading up to the European ban of kava in 2003 [18] We looked for patterns between cytotoxicity level and the presence of these potentially toxic compounds using M + H extracted ion chromatograms. Overall, only a peak corresponding to FLKC was detectable above the limit of detection (s/n.3), where the relative intensity of this peak was higher in extracts with higher toxicities. Additional experiments are necessary to understand how these compounds interact with FLKA and FLKB to produce extract cytotoxicity [17,18,26,28]. Regardless of the precise cause of cytotoxicity it is clear that tremendous variation exists in the chemical composition and resulting toxicity of commercially available kava products.

Conclusions
Kava export data show that in spite of bans and warnings, consumption of unregulated kava products appears to be increasing. Our analysis shows that the assortment of commercially available kava products vary widely in chemical composition and cytotoxicity level. Certain kava cultivars and preparation methods may produce products that vary broadly in both their toxicity and their efficacy and thus a rapid and easily applied method to characterize and classify kava products would be beneficial to the consumer. Disregarding kava and its potential use as an anxiolytic or for cancer preventive ignores the great potential societal benefits of the rational and informed medicinal use of this plant.

Solvents and reagents
HPLC grade solvents from Sigma Aldrich (St. Louis, MO, USA) were used including: acetonitrile, dimethyl sulfoxide (DMSO), formic acid, ethyl acetate, 95% ethanol, and hexanes. Reverse osmosis deionized glass distilled water was obtained in house using a Thermo Scientific Barnstead B-pure filter and Distinction water still model D4000 (Bibby Scientific Limited, Stone, Staffordshire ST15 0SA, UK). Standard kava compounds were purified from Gaia Herb (Brevard, NC, USA) commercial Kava extract. Commercial kava samples were obtained from a variety of sources (Table S1).

Kava extraction
Kava samples from 25 different sources were classified as either powder (P) or liquid (L) ( Table S1). Four extraction methods were used; methods I and II for powder samples and methods III and IV for liquid samples. The method details are as follows: Method I: 10 mL of room temperature water was added to 5 grams of powdered kava, shaken for 2 hrs, centrifuged to remove insoluble material and the supernatant evaporated to dryness and redissolved in water at a concentration of 1.5 mg of residue per mL.
Method II: the same as I, except 95% ethanol was used in place of water. Method III: 200 mL of liquid kava sample was dried in vacuo, reconstituted in 500 mL of water and then adjusted to a concentration of 1.5 mg of residue per mL with additional water. Method IV: samples were directly diluted to 1.5 mg/mL with 95% ethanol. For each extraction method four replicates per sample were prepared for analysis. Extract yield was determined gravimetrically by evaporating 500 mL of extract to dryness using a Savant model SVC-200 H SpeedVac concentrator (Farmingdale, NY, USA). Extracts were normalized to 1.5 mg/mL for experiments and stored at 4uC in the dark for no more than a week prior to LC/MS analysis. All extractions were carried out at room temperature (approximately 25uC).

Metabolic fingerprinting
Metabolic fingerprints were generated using C 18 -reversed-phase ultra-performance liquid chromatography-positive electrospray ionization-time-of-flight mass spectrometry (UPLC-ESI(+)-TOF-MS) carried out on a UPLC-TOF LCT Premier XE mass spectrometer fitted with an autosampler with a sample vial block maintained at 4uC (Acquity, Waters, Milford MA, USA). The following MS conditions were used: full scan mass scan range: 100-1000 m/z, W analyzer mode, extended dynamic range, 0.1 s scan time, desolvation temperature 350uC, desolvation nitrogen flow rate: 7.0 L/min, capillary voltage: 2900 V, sample cone voltage: 30 V, source temperature: 120uC. Separations were carried out on a reversed-phase C 18 HSS T 3 1.8 mm particle size, 2.16100 mm column (Waters). Column temperature was 50uC, solvent flow rate 0.3 mL/min, injection volume 5 mL. A 14minute gradient using mobile phases A: 0.1% formic acid in water and B: 0.1% formic acid in acetonitrile was run according to the following gradient elution profile: initial, 10%; 3 minutes, 50% B; 8 minutes, 60% B; 13 minutes, 98% B; 14 minutes, 98% B. A 7minute wash cycle was run between every sample and monitored for the absence of carryover. MassLynx version 4.1 (Waters) was used for data collection and visualization. Sample analysis order was randomized across the entire sample set.

Feature detection and multivariate statistical analysis
LC-MS files were processed using MarkerLynx version 4.1 software (Waters) for feature detection using the following parameters: mass tolerance: 0.01 Da; peak width at 5% height: 0.2 s; intensity threshold: 2000 counts; mass window: 0.05 Da; retention time window: 0.20 s. Following feature detection the feature lists were imported into Analyst version 7.5 software (Genedata, Lexington, MA, USA). Feature lists were inspected and a feature was considered to be real if it was present in greater than 75% of replicate samples with similar intensity in all replicates. Once highly confident feature lists were obtained, principle component analysis (PCA) was performed.

Absolute quantification
Absolute quantification was performed using a UPLC-single quadrupole mass spectrometer (Waters). Independent standard curves were generated for six compounds (K, DHK, M, DHM, FLKA, and FLKB) found in kava. Mixtures of pure standards were made in seven concentrations from 0.05 ppm to 100 ppm; four technical replicates were completed to account for chromatographic drift and ionization variability. After LC method optimization, standard curves were generated in selected ion recording (SIR) mode with the following retention time windows: 0-3.5 min: scan 220-700 m/z; 3.5-6.5 min: 27562 m/z; 3.5-6.5 min: 27762 m/z; 4.0-7.0 min: 23162; 4.5-7.5 min: 23362 m/z; 7.5-8.0 min scan 220-700 m/z; 8-11 min: 31562 m/z; 9-12: 28562 m/z; 12.0-14.0 min: scan 220-700 m/ z. A cone voltage of 40 V was used to disfavor non-covalent compound dimerization in ESI + mode. The liquid chromatography and column parameters are identical to those used for the metabolic fingerprinting. Standard curves were linear up to 50 ppm. Samples were analyzed using the same LC-MS method with three or four replicates in most cases, although for six samples (N, IV; X, III; Y, IV; BB, III&IV; CC, II) and four samples (J, IV; M, IV; O, IV; X, IV) only two or one replicates were suitable for the final quantification, respectively. The limit of detection was set at concentrations corresponding to a signal-to-noise ratio of 3 to 1; peaks occurring below this threshold are not detected (ND). The limit of quantification was set at a signal-to-noise ratio of 10 to 1. The MassLynx TM application manager QuanLynx TM (Waters) was used to assist with automatic integration and of this large dataset. All integrations were visually inspected and manually adjusted to ensure consistent and accurate quantification.

Kava fractionation and characterization
Fractionation of commercial kava purchased from Gaia Herb (Brevard, NC, USA) was performed as described previously [7] with normal phase silica gel chromatography generating three modalities -fraction A (hydrophilic), B (medium polarity), and C (lipophilic). Briefly, 300 mL, net weight, of kava residue was mixed with silica gel (300 g). Ethanol and water were removed by vacuum. This silica gel with adsorbed kava residue was subjected to coarse chromatographic separation using a 750-gram prepacked silica gel cartridge. The elution method was 28% ethyl acetate and 72% hexane 5 column volumes, followed by 90% ethyl acetate and 10% hexane, 4.1 column volumes, and then 35% methanol and 65% ethyl acetate, 5.5 column volumes. Different eluents were analyzed by TLC and the desired eluents were combined with solvent removed to generate fractions A, B, and C. The fractionation process was automated and monitored by the Biotag Separation System. Each individual fraction was analyzed by 1 H -NMR and HPLC to confirm the success of fractionation.

Cytotoxicity test
Cytotoxicity tests were performed on fractions A, B, & C and different kava extracts. Extracts were dried and reconstituted in DMSO at a concentration of 10 mg/mL. From these stock solutions working solutions of 150, 75, and 37.5 mg/mL were prepared from each extract. Their cytotoxicity against human lung adenocarcinoma A549 cancer cells (American Type Culture Collection CCL-185) were evaluated by following our established procedures [48]. Briefly, A549 cells were plated in a 96-well plate (2.56103 cells/well). The cells were treated with kava extracts with 0.5% DMSO in the final cell media (cells treated with media containing 0.5% DMSO served as a control). After 48 h of treatment, the relative cell viability in each well was determined by Figure 1. Histograms showing the distribution of concentrations of compounds found in commercial kava preparations. Kava samples were extracted with both 100% water (black) and 95% ethanol (red). The six compounds shown were quantified in each extract by LC-MS and the resulting concentrations in part per million are displayed histogramatically. Distributions were normalized by display on a log 2 scale. The inset y-axis scale indicates the numbers of kava samples in each bin. Measurements designated as not detected (ND) were below the limits of detection (LOD (s/n,3) for each analysis, which were typically 0.0005 to 0.001 ppm depending on variation in signal to noise from sample to sample. doi:10.1371/journal.pone.0111572.g001 Variation in Commercially Available Kava PLOS ONE | www.plosone.org using CellTiter-Blue cell viability assay kit (Promega, CA). Two biological repeats with three replicates per experiment were performed.