The approach for synergy directed fractionation developed as a result of these studies is outlined in . Crude extracts are subjected to synergy testing to identify those likely to contain synergists. Active extracts are fractionated, and each fraction is profiled using liquid chromatography-mass spectrometry (LC-MS) and again subjected to a synergy assay. LC-MS profiles are compared to bioactivity data, and statistical correlations are used to identify potential bioactive compounds (which could include both synergists and those responsible for direct activity). The process of fractionation, synergy testing, and analysis with LC-MS is repeated iteratively until sufficient pure material is obtained for structure elucidation with NMR and MS.
Figure 1 Synergy directed fractionation. Synergy testing is used to identify extracts and fractions likely to contain combinations of compounds that work together. These extracts are then fractionated, and LC-MS profiles are compared to bioassay data to identify (more ...)
The synergy directed fractionation approach is perhaps best illustrated by example. In the following section, we demonstrate its application to identify synergists from the botanical medicine goldenseal (Hydrastis canadensis
). Leaf material was chosen for this study because previous studies showed greater synergistic antimicrobial activity of goldenseal leaf extracts than goldenseal root extracts.19
The first step in the process of identifying the compounds responsible for this effect was to select an appropriate method to assay synergy. The checkerboard assay,22
which has been widely applied to study antimicrobial synergists, was chosen. To conduct the checkerboard assay, a crude H. canadensis
leaf extract was tested at a range of concentrations in combination with the known antimicrobial agent berberine (also a component of H. canadensis
). Minimum inhibitory concentrations (MIC) were measured for each combination of berberine/goldenseal extract concentrations, and an isobologram was plotted (). Wagner has presented an excellent review of the use and interpretation of isobolograms for evaluation of synergy among phytochemical compounds.2
Briefly, an isobologram is a plot where each x,y data pair represents a combination of concentrations at which a desired activity is obtained (i.e. growth of bacteria is completely inhibited). The shape of the isobologram is indicative of either synergistic, additive, or antagonistic interactions among the compounds or extracts tested. As a quantitative measure of synergy, fractional inhibitory concentrations (FIC) can also be calculated using the same data on which the isobologram is based, as described elsewhere.19
An FIC cutoff of <0.5 is usually applied to indicate synergy,23
while an antagonistic effect is characterized by FIC >4, and FIC values between 0.5 and 4 indicate no interaction. 22
In this study, the crude H. canadensis
extract demonstrated synergism when combined with berberine, as indicated by the convex shape of the isobologram () and the FIC value of 0.19 ().
Figure 2 Isobolograms for berberine, the tannin free Hydrastis canadensis extract (after liquid-liquid extraction, see Supporting Information, Figure S1), the most active fraction from the first stage of the separation (fraction 4) and the most active fraction (more ...)
MIC and FIC values (indicative of synergy) for a Hydrastis canadensis extract and fractions against wild type Staphylococcus aureus (NCTC8325-4).
Once synergy had been demonstrated for the crude H. canadensis
extract, the next step was to identify the compounds responsible for this activity. Toward this goal, the extract was fractionated (see Figure S1
provided as Supporting Information
), and the activity of each fraction in combination with berberine was evaluated. Because of limited time and materials, rather than construct a full isobologram, the activity of each fraction was measured at a single concentration (75 μ
g/mL) in combination with a range of concentrations of berberine. The MIC of berberine in the presence of the fraction was then compared to the MIC of berberine alone, and fractions were deemed “active” if they enhanced the activity of the berberine (decreased its MIC). Note that this activity does not necessarily imply synergy; a fraction could enhance the activity of berberine either
due to potentiation or to an additive antimicrobial effect. Several of the fractions collected from the first stage fractionation did, indeed, decrease the MIC of berberine (). The most active of these was fraction 4, which decreased the MIC of berberine 16-fold, from 75 μ
g/mL (berberine alone) to 4.7 μ
g/mL (berberine in combination with fraction 4).
Figure 3 Panel A shows minimum inhibitory concentration (MIC) of starting material (sm), berberine alone, and berberine in combination with 11 fractions (each at a fixed concentration of 75 μg/mL) from the first stage of separation (flash chromatography (more ...)
A limitation of the simplified approach employed for screening fractions (constant concentration of fraction in combination with a range of concentrations of berberine) is that it does not enable additive effects to be distinguished from synergistic effects. Thus, the observed activity of fraction 4 was tested further with an in-depth synergy assay. A checkerboard assay identical to that employed for the crude extract was conducted using a range of combinations of fraction 4 with berberine. The resulting isobologram () and the FIC value of 0.13 () were indicative of synergy.
Once it had been determined that fraction 4 did, indeed, contain synergist(s), LC-MS profiles were compared to the bioactivity data () to determine whether activity could be attributed to specific known or unknown compounds. LC-MS analysis revealed the presence of several ions that were not known constituents of H. canadensis in fraction 4. These ions were identified initially based on their m/z values (M - H− at 311 and 297). NMR analysis of pure isolated compounds eventually enabled their structures to be elucidated as a group of three flavonoids (1, 2, and 3), as described below.
It was possible, using LC-MS, to verify that the activity of fraction 4 was not due to the presence of the three major known alkaloids in H. canadensis
, berberine (4
), hydrastine and canadine. From , it is apparent that fractions 9, 10, and 11 contained the highest levels of berberine. The weak activity of these fractions (2-fold decrease in MIC) was attributed to this compound, and they were not investigated further. Hydrastine and canadine were present at higher concentrations in fraction 3 than in fraction 4, but fraction 4 was significantly more active. This suggested that other compounds were responsible for the activity of fraction 4, a finding that was not surprising given that hydrastine and canadine have been shown to be inactive against S. aureus
, both alone and in combination with berberine.19
Visual inspection of the data in revealed separate correlations with the presence of compounds at m/z 297 and 311. The strength of this correlation was evaluated jointly using a statistical method. A multiple correlation analysis was conducted to examine the extent of linear association between peak areas of ions 297 and 311, and the MIC of the fractions obtained from the first stage separation. A moderate correlation was observed (R = 0.77, p = 0.02), which suggests that the joint presence of ions 297 and 311 explains approximately 60% (0.772), but does not completely explain the activities observed for the fractions. Presumably, other compounds that play a role in the overall activity of the crude extract are also present. Indeed, visual inspection of the data could also have led to this conclusion, given that fractions 5 and 6 demonstrated pronounced activity (8-fold decrease in MIC of berberine) even without significant levels of berberine or of the ions with m/z 297 and 311.
Fraction 4 was subjected to a second stage of separation using flash chromatography with a hexane/ethyl acetate gradient (see fractionation scheme, Supporting Information Figure S1
). Mass spectrometry profiles were compared to MIC data (), and again the compounds with m/z
297 and 311 were present at highest concentration in the most active fraction (sub-fraction 2). Sub-fraction 2 was also subjected to the synergy assay (), and an even more pronounced synergistic enhancement of the activity of berberine was observed (FIC = 0.03). The second stage of purification yielded multiple active fractions (), which explains the weak multiple correlation observed between MIC and peak area of compounds with m/z
297 and 311 for the second stage of separation (R = 0.55, p = 0.37). This weak correlation implies that the compounds with m/z
297 and 311 do not fully account for the activity of fraction 4, and that, again, additional active compounds were present in this fraction.
Figure 4 Comparison between MIC (A) and distribution of flavonoids (B) and alkaloids (C) after the second stage of the separation (flash chromatography over silica gel with hexane:EtOAc gradient) of the Hydrastis canadensis leaf extract. Sub-fraction 2 contained (more ...)
Two rounds of preparative HPLC starting with sub-fraction 2 ultimately lead to the isolation of the flavonoids sideroxylin (1
), 8-desmethyl-sideroxylin (2
), and 6-desmethly sideroxylin (3
). These flavonoids were the same compounds observed in the LC-MS spectra at m/z
311 (sideroxylin) and 297 (isomeric 8-desmethyl-sideroxylin and 6-desmethyl-sideroxylin). The flavonoids have been reported previously as constituents of Eucalyptus spp.
and Dracaena cochinchinensis
, referred to in the reference by the name 4′,5-dihydroxy-7-methoxy-8-methylflavone).27
Ours is the first publication to report these flavonoids as constituents of H. canadensis
Other flavonoids have, however, been identified in H. canadensis9
, and, as previously mentioned, it is well known that the plant contains the alkaloid berberine (4
). Both 1
H and 13
C NMR chemical shift data (Supporting Information, S2
) were in excellent agreement with those reported for the flavonoids,24–27
and high resolution mass spectrometry measurements confirmed the molecular formulae of C18
for sideroxylin and C17
for the isomers 8-desmethyl-sideroxylin and 6-desmethyl-sideroxylin.
Although the crude extract and early stage fractions were quite soluble in the antimicrobial assay medium (Müeller-Hinton broth with 2% DMSO), the isolated flavonoids demonstrated very poor solubility (<13 μg/mL) in this same medium. As such, it was not possible to observe an influence of the pure flavonoids on the MIC of berberine against wild type S. aureus. However, activity was observed for 8-desmethyl-sideroxylin against a NorA efflux pump overexpressing strain of S. aureus (S. aureus K2378, norA++, ), which is more sensitive to inhibitors. Note that although the reported assay concentration of 8-desmethyl-sideroxylin was 75 μg/mL (), the actual concentration was likely much lower, due to the aforementioned poor flavonoid solubility. The flavonoid demonstrated no antimicrobial activity alone (MIC >300 μg/mL) against the norA++ S. aureus, but decreased the MIC of berberine 2-fold. Thus, it can be said that 8-desmethyl-sideroxylin (2) synergistically enhances, or potentiates, the antimicrobial activity of berberine against norA++ S. aureus.
MIC of berberine alone and in combination with 8-desmethyl-sideroxylin (2) against S. aureus K2378 (a norA++ strain). Reserpine served as a positive control.
Previous literature has shown flavonoids to act as efflux pump inhibitors,20,29–30
and our group has demonstrated efflux pump inhibitory activity for goldenseal leaf extracts.19
Thus, we hypothesized efflux pump inhibition as a likely mode of action for the H. canadensis
flavonoids. To test this hypothesis, the activity of each of the three individual flavonoids was evaluated with an ethidium bromide efflux assay (). Ethidium bromide is a substrate of the S. aureus
efflux pump NorA, and fluoresces strongly (due to intercalation with DNA) inside bacterial cells.20
With the ethidium bromide efflux assay, cells are loaded with ethidium bromide and fluorescence is monitored over time. A decrease in fluorescence indicates efflux of ethidium bromide from the cells, and in the presence of an efflux pump inhibitor, fluorescence should decrease more slowly. The ethidium bromide efflux assay has the advantages of being more sensitive than the checkerboard synergy assay, and also having a greater tolerance for DMSO (10% DMSO was used in this assay, as compared to 2% DMSO in the checkerboard assays). These features enabled the solubility problems with the pure flavonoids to be circumvented when using the efflux assay.
Figure 5 The three flavonoids from Hydrastis canadensis inhibit the NorA efflux pump of Staphylococcus aureus. Decrease in fluorescence over time is due to efflux of ethidium bromide, which is blocked by the positive control (CCCP) and all three flavonoids, sideroxylin (more ...)
shows the results of the ethidium bromide efflux assay with wild type S. aureus (NCTC 8325-4). The positive control, carbonyl cyanide m-chlorophenylhydrazone (CCCP), and compounds 1, 2, and 3 inhibited efflux of ethidium bromide. Significant differences between vehicle control (10% DMSO) and treatment with sideroxylin (1) (p = 0.02), 8-desmethyl-sideroxylin (2) (p = 0.0002), 6-desmethyl-sideroxylin (3) (p = 0.002), and CCCP (p = 0.0002) were observed at t = 300 s. When the same experiment was repeated using the norA deletion mutant S. aureus (K1758), neither CCCP nor any of the flavonoids significantly inhibited efflux of ethidium bromide (). The results demonstrate that all three flavonoids from S. aureus inhibit the NorA efflux pump of S. aureus.
Although all three flavonoids demonstrated some efflux pump inhibitory activity, they differed in potency (). Sideroxylin (1), the dimethyl analog, was a less effective inhibitor than either of the monomethyl analogs, 8-desmethyl-sideroxylin (2) and 6-desmethyl-sideroxylin (3). It appears that substitution on both the 6 and 8 positions decreases efflux pump inhibitory activity.
A relevant question to the quality control of dietary supplements from H. canadensis
aerial portions is whether the flavonoids and alkaloids are present at detectable levels in such preparations. To address this question, ethanolic extracts were analyzed that had been prepared using a method based on that employed in the manufacture of dietary supplements.31
Both flavonoids and alkaloids were detected in the ethanolic extracts. Retention times and CID fragmentation patterns for these compounds (see Supporting Information, Table S2
) matched those of standards. In agreement with previous literature, the alkaloids were present at higher levels in extracts prepared from goldenseal roots than leaves.18
Conversely, the flavonoids were detected only at very low levels in the root/rhizome extracts, but at much higher levels in the leaf extracts (). Goldenseal leaf extracts contained an average of 1.9 ± 0.4 mM 8-desmethyl sideroxylin/6-desmethyl-sideroxylin (quantified as a mixture) and 73.3 ± 0.9 μ
M sideroxylin. Levels in the root extracts were over 50 times lower. Notably, the same leaf extracts for which data is presented in have been shown previously to have efflux pump inhibitory activity,19
which is likely attributable, at least in part, to the presence of flavonoids 1
, and 3
Table 3 Concentrations (± SE, N = 6) of alkaloids (berberine (4), hydrastine, canadine) and flavonoids [sideroxylin (1), 8-demethy-sideroxylin (2), and 6-desmethyl-sideroxylin (3)] in extracts from the roots/rhizomes and leaves of Hydrastis canadensis (more ...)
The finding that goldenseal leaf extracts have higher levels of synergists while root extracts contain higher levels of alkaloids suggests the potential benefit of using a mixture of root and leaf material in the production of dietary supplements from goldenseal. Further studies would, however, be needed to evaluate the safety and efficacy of goldenseal leaf extracts in vivo. If goldenseal leaf material were shown to be efficacious, this could add value to cultivated goldenseal crops. In addition, the use of goldenseal leaf material has the added advantage of reducing impact on wild goldenseal populations. Goldenseal leaves can be harvested in the fall, after the berries have dropped, without killing the plants. Sustainable production of goldenseal would be particularly desirable given that it is listed as threatened by CITES in much of its native habitat.32
In terms of the broader relevance of this study, it has been shown that synergy directed fractionation can be employed to identify both compounds with direct antimicrobial activity (berberine) and efflux pump inhibitors that synergistically enhance berberine’s activity (the three flavonoids) from a complex botanical extract. The results highlight three important elements that distinguish synergy directed fractionation from bioactivity guided fractionation. First, the application of synergy testing to the crude extract made it possible to determine that it contained synergists even prior to fractionation. Second, integration of synergy testing with the fractionation process allowed synergists to be tracked as they were purified. This process ultimately facilitated the isolation of synergists that possess no inherent antimicrobial activity, compounds that may have been missed in inactive fractions with traditional bioactivity guided fractionation. Third, comparison of LC-MS data and biological activity data after each stage of separation allowed potential active compounds to be identified and tracked throughout the isolation process. This establishes an important linkage between biological activity of the crude extract and the presence of specific active compounds. In the example presented here, it was possible based on LC-MS profiles to identify ions corresponding to the active flavonoids even after only one stage of separation. Importantly, it was not necessary to have prior knowledge of the identities of the compounds detected in the active fractions to correlate their presence with biological activity. These compounds were initially identified based solely on characteristic m/z values, and it was only after their likely role in the biological activity of the extract had been established that they were isolated and their identities determined via NMR.
Visual inspection and statistical analyses of the data after the first and second stages of separation suggests the presence of multiple active compounds, including some that have not yet been identified. By no means does this finding negate the importance of the flavonoids and alkaloids in the activity of H. canadensis. Rather, the results suggest that the antimicrobial activity of the crude H. canadensis extract is due to multiple compounds, including antimicrobial alkaloids, flavonoids that synergistically enhance the antimicrobial activity of the alkaloids, and several additional compounds whose identities and modes of action have yet to be determined. This finding is consistent with the claims often made for botanical medicines; that their activity results from the combined (and perhaps synergistic) action of an array of chemically diverse constituents.