Interaction Metabolomics to Discover Synergists in Natural Product Mixtures

Abstract

Mass spectrometry metabolomics has become increasingly popular as an integral aspect of studies to identify active compounds from natural product mixtures. Classical metabolomics data analysis approaches do not consider the possibility that interactions (such as synergy) could occur between mixture components. With this study, we developed "interaction metabolomics" to overcome this limitation. The innovation of interaction metabolomics is the inclusion of compound interaction terms (CITs), which are calculated as the product of the intensities of each pair of features (detected ions) in the data matrix. Herein, we tested the utility of interaction metabolomics by spiking known concentrations of an antimicrobial compound (berberine) and a synergist (piperine) into a set of inactive matrices. We measured the antimicrobial activity for each of the resulting mixtures against Staphylococcus aureus and analyzed the mixtures with liquid chromatography coupled to high-resolution mass spectrometry. When the data set was processed without CITs (classical metabolomics), statistical analysis yielded a pattern of false positives. However, interaction metabolomics correctly identified berberine and piperine as the compounds responsible for the synergistic activity. To further validate the interaction metabolomics approach, we prepared mixtures from extracts of goldenseal (Hydrastis canadensis) and habañero pepper (Capsicum chinense) and correctly correlated synergistic activity of these mixtures to the combined action of berberine and several capsaicinoids. Our results demonstrate the utility of a conceptually new approach for identifying synergists in mixtures that may be useful for applications in natural products research and other research areas that require comprehensive mixture analysis.

Figure 1

Checkerboard assay results of berberine and piperine combinations againstS. aureus. (A, B) The dose–response curves of berberine (1) and piperine (2) against S. aureus, respectively. (C) A dose–response curve of berberine combined with different concentrations of piperine (shown in different colors). Without added piperine, the MIC of berberine is 150 μg/mL. The addition of piperine reduces the MIC of berberine. For example, the MIC of berberine is 9.38 μg/mL in the presence of 50 μg/mL piperine. (D) A hyperbolic isobologram, indicating synergy. Additionally, the ∑FIC value (eq 1) of berberine in the presence of 50 μg/mL piperine is 0.19, signifying synergy.

Figure 2

Experimental workflow for preparation and analysis of spiked fractions. A simulated extract was prepared by mixing 21 natural products that did not demonstrate antimicrobial activity against S. aureus, alone or in combination with berberine and piperine. This mixture was fractionated, and the pooled fractions were used as background matrices to create spiked fractions containing known amounts of berberine (antimicrobial) and piperine (synergist). Untargeted metabolomics data were collected with ultraperformance LC-MS to obtain a metabolomics data set. Antimicrobial activity was evaluated for all fractions against S. aureus to obtain a biological data set.

Figure 3

Comparison of biological and chemical data demonstrates the utility of the CIT. The antimicrobial activity (% inhibition of S. aureus growth) was strongly correlated with standardized abundance for the fractions spiked with berberine (M01-M08, panels A and B). For the mixtures spiked with berberine and piperine (M09-M17), neither berberine (C) nor piperine (E) abundance tracks with biological activity. This lack of correlation is shown with poor linearity of regression plots for % inhibition vs peak area of berberine (D) and piperine (F). The CIT obtained by multiplying the peak area for piperine with the peak area for berberine (eq 3) tracks with biological activity for the berberine-piperine mixtures (G). The linear relationship between the % inhibition and the CIT is demonstrated in panel H. To generate these plots, piperine abundance was measured as the peak area of the selected ion for the [M + H]⁺ ion of piperine detected at m/z 286.1426, while the peak area of berberine was measured as the peak area of the M⁺ ion detected at m/z 336.1217. The peak areas for these ions were standardized, as shown by eq 5.

Figure 4

Two possible metabolomics workflows for data analysis: (A) classical metabolomics and (B) interaction metabolomics. The two workflows both start with a biological data set and a metabolomics data set (in this case LC-MS data obtained by analysis each of the mixtures individually). The values from these biological and metabolomics data sets are compiled in one of two data matrices (Figure S8). The data matrix for the interaction workflow differs in the inclusion of the CITs. The values in the data matrices are then standardized (eq 5), and multivariate statistical analysis is conducted, resulting in the prediction of putative antimicrobials (classical metabolomics) or putative synergists (interaction metabolomics).

Figure 5

Comparison of selectivity ratio plots for the spiked fractions using classical metabolomics (A, B) and interaction metabolomics (C, D). The five CITs in (C) are the products of berberine and the five piperine adducts: protonated species [M + H]⁺, sodiated species [M + Na]⁺, proton bound dimer [2M+H]⁺, sodium bound dimer [2M+Na]⁺, and a sodiated acetonitrile cluster [M+ACN+Na]⁺. Specifically, in Panel C, 13 × 23 = piperine [M + H]⁺ × berberine M⁺, 19 × 23 = piperine [M + Na]⁺ × berberine M⁺, 23 × 24 = berberine M⁺ × piperine [M+ACN+Na]⁺, 23 × 32 = berberine M⁺ × piperine [2M+H]⁺, and 23 × 33 = berberine M⁺ × piperine [2M+Na]⁺. In panel D, 21 × 22 = berberine M⁺ × [M + H]⁺ of putative 7,8-dihydroberberine. The total number of features in the data set were 17, 32, 528, and 153 for panels A, B, C, and D, respectively. Prior to multivariate statistical analysis, all metabolomics data were filtered based on the requirement that they demonstrate consistent peak area across all replicate analyses and the requirement that the feature area vary by >0.01% across all samples in the mixture (see Experimental Section).

Figure 6

Workflow used to prepare fractions from Hydrastis canadensis and Capsicum chinense (A) and procedure for mixing fractions (B). To prepare the subfractions of H. canadensis, the H. canadensis partition highest in berberine (HC-aq) was separated into three fractions with flash chromatography. The fraction with highest berberine content (HCF2) was further fractionated into six subfractions using preparative HPLC, and of these, HCF2–6 had the strongest antimicrobial activity (32% inhibition of S. aureus at 50 μg/mL, Figure S11A) and the highest content of berberine (Figure S12). To prepare the subfractions of C. chinense, the partition with highest content of capsaicin (CC-ch) was selected and separated into 12 fractions using flash chromatography. The fraction with the highest capsaicin content (CCF4) was then fractioned into five subfractions, CCF4-1, CCF4-2, CCF4-3, CCF-4, and CCF-5, using preparative HPLC (Figure S13). Mixtures were prepared by combining a sub-MIC concentration (50 μg/mL) of the most strongly antimicrobial subfraction from Hydrastis canadensis (HCF2–6) with all of the subfractions from the C. chinense (CCF4-1 through CCF4-5), each at 100 μg/mL.

Figure 7

Selectivity ratio plot generated with classical metabolomics (A) and interaction metabolomics (B) using the metabolomics data and antimicrobial data obtained from the H. canadensis and C. chinense mixtures prepared as shown in Figure 6. For classical metabolomics (A) the predominant feature in the selectivity ratio plot corresponds to protonated berberine (the M⁺ ion with m/z 336.1230). With interaction metabolomics (B), synergists are also detected. Of the 17 high selectivity ratio features or CITs highlighted in red, 16 could be assigned either to berberine alone or to berberine interacting with one of four putative capsaicinoids.