Promotion of quality standard of herbal medicine by constituent removing and adding

To identify major active constituents and measure their levels in a typical medicinal herb–Rhizoma coptidis, we applied the concept of removing and adding, taking inspiration from functional genetic methods. As this herb has bacteriostatic properties and is used to treat bacterial diarrhea, we examined the effects of individual constituents (berberine, palmatine, coptisine, epiberberine, jateorrhizine and columbamine) on the growth of Shigella dysenteriae with microcalorimetry. The removing and adding procedures revealed that berberine and coptisine were the main antibacterial constituents of R. coptidis, with bacteriostatic activities of 54.10% and 39.75%, respectively. The relative levels of berberine and coptisine in R. coptidis were 8.08%–31.92% and 4.05%–14.45%, respectively. On the basis of whole effect, the method of constituents removing and adding, coupled with a bioassay, is a useful strategy to identify the active constituents and measure their levels in herbal medicines, which may provide reference to other natural products.

Rhizoma coptidis (R. coptidis), an important herbal medicine to treat bacterial diarrhea, constipation, diabetes, and other major diseases in Asia, Africa, and Latin America for a long history 32,34 ( Supplementary Fig. S2), was studied in this research. There are a lot of protoberberine alkaloids in R. coptidis, of which the structure belongs to isoquinoline alkaloids. The structural differences of those alkaloids are mainly caused by different groups bound to C2, C3, C9 and C10 (Supplementary Fig. S3). But their contributions to the bacteriostatic activity of R. coptidis and the content levels have not been clearly specified. Thus, we put forward the strategy of constituent removing and adding, hoping to discover active constituents of R. coptidis inhibiting Shigella dysenteriae (S. dysenteriae) and their content limits. Specifically, we first removed the targeted constituents from R. coptidis to evaluate the inhibitory effects of the targeted constituent and its negative sample (samples lacking the targeted constituent) on the activity of S. dysenteriae, with unmodified R. coptidis extract as a reference sample. After confirming the identities of the active constituents, the targeted constituents were added into the negative samples individually to measure their bacteriostatic effects. We believe that the difficulty of characterizing the active constituents and measuring their levels in herbs can be overcome by applying the removing and adding strategy together with a bioassay. This method can provide an important reference for the establishment of QS related to the clinical efficacy. The experimental strategy used in this study is shown in Fig. 1.

Results
Constituent removing and chemical identification. A preparative thin layer chromatography plate (PTLCP) was used to achieve constituent removal. Each band on the plate corresponded to one constituent and was regarded as a positive sample (namely targeted constituent marked by letter M 1 ). The rest of bands was defined as the corresponding negative sample (marked by letter M 2 ).
Next, we used Ultra-performance liquid chromatography (UPLC) and quadrupole time-of-flight mass spectrometry (Q-TOF MS) to identify the constituent of each band, and to determine whether there were residual compounds in the corresponding negative sample. Using this method, the constituents in the positive samples were preliminarily identified ( Supplementary Fig. S4) as berberine (BER), palmatine (PAL), coptisine (COP), epiberberine (EPI), jateorrhizine (JAT), and columbamine (COL), respectively. The levels of the removed constituents in the negative samples were nearly undetectable ( Supplementary Fig. S4). Furthermore, the molecular structures of the removed constituents were not damaged by the removal procedure ( Supplementary Fig. S5).
Identification of the active constituents. The growth and metabolism of living organisms are accompanied by heat/energy production, which can be affected by pathological changes or the action of drugs. Therefore, it is possible to evaluate changes in microbial heat production in the presence or absence of different drugs using microcalorimetry. Accordingly, we determined the bioactivity of R. coptidis extract, removed samples, and added samples in terms of bacteriostasis.
The normal growth thermogenic curve for S. dysenteriae at 37uC is shown in Fig. 2a. The heat flow power-time (HFP-t) curve showed that the S. dysenteriae metabolic profile included two main stages (stages 1 and 2) and five phases, (lag phase where P 0 and P t represent the heat flow power at time 0 or time (min), respectively. To test the reliability of the microcalorimetry, we repeated the experiment on eight occasions in untreated bacteria and obtained good reproducibility. We then quantified the following thermokinetic parameters from the HFP-t curves in the presence of difference concentrations of the samples: p 1 , p 2 , t 1 and t 2 ( Table 1). PCA revealed that k 2 and t 2 explained 87% of the variation of samples, including the R. coptidis extract, removed samples, and added samples. Therefore, we focused on parameters k 2 and t 2 in this study.
Figures 2b, c illustrate the effects of R. coptidis (0.8 mg/mL) extract, removed constituents, and corresponding negative samples on the HFP-t curves of S. dysenteriae. We found that, compared with the control and the R. coptidis extract, the kinetics of the removed and negative samples showed marked variation (Supplementary  Table S1).
Next, we calculated the inhibition ratio I as equation (2) and the relative inhibition ratio RI as equation (3) to quantify the contributions of the individual constituent to the bacteriostatic activity of R. coptidis extract. As shown in Fig. 2d, BER, COP, EPI, PAL, and JAT 1 COL removed inhibited the growth and metabolism of S. dysenteriae; of which BER and COP had the greatest effects. The bacteriostatic activities of BER 2 and COP 2 were significantly lower than that of R. coptidis extract (P , 0.01). Furthermore, the bacteriostatic activities of EPI 2 , PAL 2 , and (JAT 1 COL) 2 were not significantly different to that of R. coptidis extract. Thus, BER and COP appear to be the main bacteriostatic constituents of R. coptidis extract with contributions to the bacteriostatic activity of 54.10% and 39.75%, respectively (Fig. 2d).
where K 2c is the growth rate constant of the second exponential growth phase of S. dysenteriae in the culture medium alone; k 2s is the growth rate constant of the second exponential growth phase of S. dysenteriae exposed to the test samples; I s is the inhibition ratio of samples under evaluation; I e is the inhibition ratio of R. coptidis extract (reference); and RI is the relative inhibition ratio.
Constituent adding and chemical identification. After confirming that BER and COP were the main bioactive constituents of R. coptidis, we added BER to its negative sample (BER 2 ) to final concentrations of 0, 15, 45, 60, 80, and 120 mg/mL, which was equivalent to the relative content of BER in R. coptidis extract was Levels of the active constituents. The bacteriostatic activity of samples adding BER or COP was assessed by microcalorimetry ( Fig. 3c,d), and I and RI were calculated (Tables 1 and 2). The relationship between RI and BER or COP concentrations are shown in Fig. 4. To determine the correlation between the concentrations of active constituents in added samples and I, we calculated the change Table 1 | Thermokinetic characteristics of R. coptidis (200 mg/mL) and samples containing different concentrations of BER on S. dysenteriae growth at 37uC (n 5 3, Mean 6 SD)  in I induced by the each concentration of active constituents (P) using equation (4): where I n is the inhibition ratio of the negative sample lacking the targeted constituent; I i is the inhibition ratio of the negative sample following the targeted constituent added; and W is the corresponding total concentration of the targeted constituent causing I i .
As illustrated in Fig. 4, the relative potencies of COP (i.e., P COP ) and BER (i.e., P BER ) were similar (red curves in Fig. 4). For both constituents, their relative potencies increased with increasing weight. However, the potency started to decrease when the weight passed a threshold level. Therefore, the peak values in the P COP and P BER curves (e and f in Fig. 4) are likely to show the greatest potencies. Therefore, the concentrations of COP and BER that showed the greatest efficiency were 32 mg/mL and 80 mg/mL, respectively. We calculated that the greatest efficiency of COP and BER is at the relative concentrations 14.45% and 31.92%, respectively.
The dose-response curves in Fig. 4 (blue curves) revealed some differences in I COP and I BER . Points a and b in Fig. 4 correspond to the RIs of COP and BER, respectively. These values differed because the corresponding negative samples contained other constituents. Points c and d in Fig. 4 correspond to the RIs of 8 mg/mL COP and 15 mg/mL BER, respectively. At concentrations of COP and BER exceeding 8 mg/mL and 15 mg/mL, respectively, the RIs increased rapidly, and there was a significant difference between I i and I n . We also found that the relative compositions of COP and BER which showed the best efficiency were 4.05% and 8.08%, respectively (Tables 1 and 2). The RI of COP did not increase markedly with further increases in concentrations after point g. Thus, the concentration of 32 mg/mL, corresponding to point g, might be the maximum dose of COP. By contrast, because there was no apparent plateau for I BER , we could not determine the maximum dose of BER. Although this finding might be due to synergistic effects of BER combined with other constituents of the negative samples, further studies are needed to confirm this possibility. By comparing the I BER and P BER curves, we found that P BER decreased sharply after point f, but I BER did not increase substantially. Accordingly, we speculate that the concentration of 80 mg/ mL, which corresponds to point f, could be the maximum dose limit of BER. By comparing the relative compositions of the knocked in constituents (Table 1), we found that the relative composition of BER with the greatest efficiency was 31.92%.

Discussion
Currently, the choice of QC indicators for herbal medicines is often subjective without forceful evidences [12][13][14] . Inspired by the discovery of functional genes, we put forward the strategy of constituent removing and adding according to the dual attributes of genetic information (genes) and chemical information (constituents) of organisms, hoping to promote the QC of herbal medicines. With this method, we found that the main bioactive constituents in R. coptidis that inhibited the growth of S. dysenteriae were BER and COP, with contributions to the bacteriostatic activity of 54.10% and 39.75%, respectively. By adding BER and COP into their negative samples, we found that the relative concentrations of BER and COP were 8.08%-31.92% and 4.05%-14.45%, respectively. We also found that BER alone did not fully account for the bacteriostatic activity of R. coptidis, although BER preparations are now widely used in clinical settings.
Although our strategy was inspired by gene knockout and knockin, there are fundamental differences between these methods. In particular, gene knockout and knock-in involves gene recombination, and changes to the expression of genes through knockout, knock down, and silencing, which has dynamic features and can be expanded [19][20][21][22][23] . In terms of removing and adding of the constituents of an herbal medicine, there are five technical issues that need to be considered: (1) What constituent is being removed? (2) Was the removed constituent a single compound? (3) Does removal affect the structure of the targeted constituent? (4) How to add the constituent? (5) Are there any sensitive and high-throughput methods that can be used to assess the bioactivity? To deal with those issues in the study of constituent removing and adding of herbs, we used the convenient and reliable PTLCP to remove the bioactive constituents capable of inhibiting the growth of S. dysenteriae. Although JAT and    COL were not separated completely, the result showed that the combination of them contributed little to the bacteriostatic activity of R. coptidis extract, which meant that the two constituents were not mainly active ones. Additionally, we performed HPLC and MS-MS to confirm that the removed constituents were not damaged, and there were almost no residual constituents in the negative samples. We also verified that the active constituents were added correctly. By using PTLCP, we were able to prepare sufficient quantities of samples. Although there are many methods available to assess bacteriostatic activity, we used a sensitive and high-throughput method called microcalorimetry to assess the bioactivity of the removed and added constituents.
A common characteristic of complex living organisms, such as herbal medicines, is that all of the constituents work together to produce an effect that is greater than the effects of the individual constituents. Unlike the traditional method of separating and then testing the activity of individual phytochemicals, we tried to avoid the limitations of reductionism in studies of complex systems without considering the interactions among constituents. Instead, we removed the targeted constituents from a complex system and then added them into the system to investigate the activity of samples. In this way, we ensure our conclusions can be used to establish a QS for herbs that is suitable for the clinical use and has guaranteed the clinical efficacy. Although there is a limitation of the strategy for its dependence on the separation of constituents, the combination of multiple techniques such as preparative high performance liquid chromatography and high speed counter current chromatography could be used to remove the targeted constituent [35][36][37] . Meanwhile, with the development of further researches on the constituent separation of herbal medicines, we believe that the strategy will also be improved better. Further studies that evaluate the synergistic and antagonistic properties of the bioactive constituents of herbs would aid the development of multicomponent drugs.

Methods
Materials. R. coptidis (Coptis chinensis Franch.) samples of about 5 years old were obtained using Good Agricultural Practice from Shizhu county, Chongqing province, China, in October 2010. All of the samples satisfied the standard for R. coptidis in the Chinese Pharmacopoeia and the Japanese Pharmacopoeia.
Reference compounds, including BER, PAL, COP, EPI, JAT, and COL were obtained from the National Institutes for Food and Drug Control (NIFDC) (Beijing, China); their purities were not less than 99.0% by high-performance liquid chromatography (HPLC).
S. dysenteriae (CMCC B 51252) was provided by the China Center for Type Culture Collection (Wuhan University, Wuhan, China). S. dysenteriae is the main pathogen causing dysentery. R. coptidis cures dysentery by inhibiting such pathogens. S. dysenteriae is frequently used to study the bioactivity of drugs used to treat diarrhea. Luria-Bertani (L.B.) culture medium was used in the bioassay and contained 5 g yeast extract, 5 g sodium chloride, and 10 g peptone in 1000 mL (pH 5 7.2-7.4). The culture medium was sterilized at 121uC under 0.1 MPa for 30 min, and stored at 4uC until required.
Acetonitrile and methanol (chromatographic grade) were purchased from Fisher Chemicals (Pittsburgh, PA). Water was purified using a Milli-Q Water Purification System (Millipore, Bedford, MA). Silica Gel G preparative thin layer chromatography plates (PTLCP, 20 3 20 cm wide, 1 mm thick) were purchased from Yantai Chemical Industry Research Institute (Shandong, China). All other chemicals were purchased from Beijing Chemical Factory (Beijing, China) and were of analytical grade.
Preparation of the Standard Solutions and the R. coptidis Extract. Standard solutions containing BER, PAL, COP, EPI, JAT, and COL were prepared by accurately weighing specific amounts of each substance and dissolving them in methanol in a volumetric flask. The concentrations of the standard solutions were as follows: BER 0.155 mg/mL, PAL 0.045 mg/mL, COP 0.032 mg/mL, EPI 0.035 mg/mL, JAT 0.016 mg/mL, and COL 0.018 mg/mL. The R. coptidis extract was prepared as follows. The herb was first dried at room temperature and milled into a powder. Then, about 50 g of the powder was transferred to a flask containing 500 mL of purified water and mixed for 30 min. Reflux extraction was performed twice for 1.0 h each. The resulting solution was filtered while it was still hot and the filtrate was collected. Finally, the filtrate was evaporated to dryness and the solid residue was dried in a desiccator.
Constituent removing. A silica Gel G PTLCP was used to separate and obtain the constituents of R. coptidis. The water extract of R. coptidis, and standard solutions of BER, PAL, COP, EPI, JAT, and COL were spotted onto the same plate, and developed in a mixture of toluene, ethylacetate, methanol, isopropanol and ammonia water (65351.551.550.5, v/v). After observing the plate under daylight, the bands corresponding to the relative retention values of each reference compound were retrieved. These procedures were repeated. Then, the merged silica chips were extracted with methanol and the suspensions were filtered and centrifuged. The supernatants were collected and evaporated to dryness for further chemical analysis and for use in bioassays to identify the active constituents in R. coptidis.
Constituent adding. After identifying the bioactive constituents of R. coptidis, we applied the added procedure. In this phase, the active constituents were dissolved in L.B. culture medium to the required concentrations. Then, the dissolved constituent was added into the negative sample (prepared in L.B. culture medium) without the corresponding constituent. Then, we performed chemical analyses to characterize the constituents. Finally, microcalorimetry was performed to determine the bioactivity of different amounts of each active constituent to determine their levels in the herbal extract.
Chemical analysis. HPLC was performed to determine the levels of the active constituents in R. coptidis extract, evaluate the changes in the chemical constituents following removing and adding, as follows. HPLC was performed using an Agilent 1200 HPLC system (Agilent Technologies, Santa Clara, CA). Chromatographic separation and detection of samples were performed on a Kromasil TM C 18 column (250 mm 3 4.6 mm, 5 mm) at a column temperature of 30uC with a flow rate of 0.6 mL/min using a solution of acetonitrile/water (0.05 mol/L potassium dihydrogen phosphate) (50550, v/v). Then, sodium dodecyl sulfate was added to a concentration of 0.4 g/100 mL of solution (pH 5 4.0). The standard solutions, R. coptidis extract, active constituents obtained by PTLCP separation, and the added samples were filtered through a 0.22 mM Millipore membrane (Carrigtwohill, Co. Cork, Ireland) and injected into the HPLC system for chemical analysis.
Bioactivity assay. The bioactivity experiments were performed using the ampoule method at 37uC with a 3114/3236 TAM air isothermal microcalorimeter (Thermometric AB, Jarfalla, Sweden). S. dysenteriae were inoculated into 100 mL of L.B. culture medium at an initial density of 1 3 10 6 colony forming units (CFU) per mL. Then, 10 mL of the bacterial suspension was added to sterilized 20-mL glass ampoules. The R. coptidis extract, removed samples, and added samples were then added to the bacterial suspension. Each ampoule was sealed and placed in an eightchannel calorimeter. When the temperature of the ampoules reached 37uC, heat flow power-time (HFP-t) curves were recorded for each sample until the values returned to baseline. Each experiment was repeated three times. All data were continuously recorded using PicoLog TC-80 software (TA Corporation, New Castle, USA).
Microcalorimetry data are repeatable and the method provides real-time, online, dynamic information to characterize the amount of heat produced by organisms. We measured the following thermokinetic parameters: the growth rate constants of the first and second exponential phases (k 1 and k 2 ), the heat flow power (HFP) of the first and the second highest peaks (p 1 and p 2 ), and the appearance time of the first and second highest peaks (t 1 and t 2 ) from the HFP-t curves.
Data analysis. Quantitative thermokinetic parameters were analyzed using Origin 8.5 software (OriginLab Company, Northampton, USA) to plot the HFP-t curves, and calculate the thermokinetic parameters.
Principal component analysis (PCA) is a data reduction technique that can be used to extract data, remove redundant information, highlight hidden features, and visualize the relationships among numerous variables with a small number of underlying factors (principal components or PCs) without losing crucial information 38 . Therefore, we performed PCA to identify the main parameters derived from the HFP-t curves.
We also applied analysis of variance (ANOVA) to test for differences in the bioactivity of different samples, and identify the active components. This procedure calculates whether there is a significant difference between the observed value (i.e., F value) and the corresponding probability (P value).