Antibacterial evaluation of Salvia miltiorrhizae on Escherichia coli by microcalorimetry coupled with chemometrics

For seeking novel antibacterial agents with high efficacy and low toxicity to deal with drug resistance, the effects of Salvia miltiorrhizae from various sources on Escherichia coli were evaluated by microcalorimetry coupled with chemometrics. Firstly, the heat-flow power-time curves of E. coli growth affected by different S. miltiorrhizae samples were recorded. Then, some crucial quantitative thermo-kinetic parameters including growth rate constant, heat-flow power and heat output, etc. were obtained from theses curves and were further investigated by some powerful chemometric techniques including similarity analysis, multivariate analysis of variance, hierarchical clustering analysis and principle component analysis. By analyzing the principle parameters, growth rate constant of the second exponential phase (k 2) and the heat-flow output powers of the second highest peak (P 2), together with the derived parameter inhibitory ratio (I, %), it could be quickly concluded that the tested S. miltiorrhizae samples from different sources in China exhibited strong antibacterial effects on E. coli and the samples from Beijing city exhibited the strongest anti-E. coli effects, which might be used as novel and underlying antibacterial candidates for the resistance of E. coli to the existing drugs in practice. This study provides a useful tool and helpful idea to accurately and rapidly evaluate the antibacterial effects of some complex matrices, offering some references for exploring new antibacterial agents.


Introduction
In recent decades, more and more attentions have been paid on Traditional Chinese Medicines (TCMs) because of their complementary therapeutic efficacy to Western medicines, and their abilities to solve some primary problems that have not yet been solved by traditional therapy, such as resistance of some microbes to the existing antibacterial agents which has led to increasing challenges for doctors and researchers, as well as has become an increasingly important and pressing global attention As a well-known and important medicinal plant derived from the dried root of Salvia miltiorrhizae Bge., Salvia miltiorrhizae (Danshen in Chinese) has been officially recorded in Chinese pharmacopoeia for the treatment of cardiovascular diseases, inflammation, mental and liver diseases (Guo et al. 2014;Zhang et al. 2016;Zhou et al. 2005). The main bioactive components of S. miltiorrhizae such as hydrophilic phenolic acids and lipophilic diterpenoid tanshinone have been proven to express widely antibacterial activities with potential applications in medicinal industry (Wang et al. 2007;Zhao et al. 2011).
Microcalorimetry has been used successfully to evaluate the antibacterial effects of TCMs with high sensitivity, accuracy, and low time-consuming (Zhao et al. 2014). During the metabolic growth processes, a flow of thermal effect is generated, which can be recorded by a microcalorimeter and is directly related to an increase or decrease in the power release by different sources. For recording the evolution of energetic intensity, the microcalorimetric tool shows special advantages compared with conventional techniques of biological investigation, such as disk diffusion method (Wu et al. 2013). By means of analyzing the heat-flow power (HFP)-time curves of microbial metabolic growth recorded by the microcalorimeter, the thermokinetic parameters such as growth rate constant (k), heat output (Q) and heat-flow power (P) can be obtained, and the effects of other substances on microbes can be further well and effectively evaluated and compared by analyzing the changes of these parameters (Braissant et al. 2013(Braissant et al. , 2015Chen et al. 2013;Zhao et al. 2015) in combination with some powerful chemometric approaches such as similarity analysis (SA), multivariate analysis of variance (MANOVA), hierarchical clustering analysis (HCA) and principle component analysis (PCA).
To the best of our knowledge, antibacterial evaluation and comparison of S. miltiorrhizae from various sources on the growth of various microbe sources by using the microcalorimetric technique has not been reported. The purpose of this study was to determine the antibacterial activities of S. miltiorrhizae on E. coli by using microcalorimetry coupled with some helpful chemometric methods including SA, MANOVA, HCA and PCA. The results have shown that the microcalorimetric technique was a potential and powerful tool to effectively investigate and evaluate the antibacterial activities of TCMs and S. miltiorrhizae with high efficacy and low toxicity can be used as a novel and underlying antibacterial candidate for the resistance of E. coli to the existing drugs.

Plant materials and chemicals
Thirty-two batches of S. miltiorrhizae samples (Table 1) with different sources were purchased or collected from various places in China and were identified by Prof. Yulin Lin (Institute of Medicinal Plant Development, Chinese Academy of Medical Sciences and Peking Union Medical College, PR China), which were labeled as XB-1, XB-2… XB-17 from Guizhou province, FC-1, FC-2…FC-6 from Sichuan province, SD-1, SD-2…SD-4 from Shandong province, HN-1, HN-2, HN-3 from Henan province, and T-1, H-1 from Beijing city. All samples were sterilized and collected in polyethylene bags, subsequently stored at −20 °C before test. Preliminarily-purified water by a Milli-Q water purification system (Millipore, Bedford, MA) was used for preparing the sample solutions. All other chemicals used were of analytical grade and available locally.

Bacterial strains and culture media
The strain of E. coli (CCTCC AB91112) was provided by China Center for Type Culture Collection, Wuhan University, Wuhan, P.R. China. Firstly, E. coli were inoculated into 25 mL broth culture medium which contained 10 g peptone, 6 g beef extract and 5 g NaCl in 1000 mL purified water (pH7.0-7.2) and was sterilized by autoclaving at 0.1 MPa and 121 °C for 30 min. Then, the 100 mL widemouthed glass bottle containing inoculated medium was incubated in a ZWFR-200 shaker (Shanghai, China) for 8 h at 37 °C at the rotation speed of 110 rpm. The flask was sealed up with parafilm for limiting the quantity of oxygen because E. coli is a kind of facultative anaerobe. After incubation, the bacteria were transferred into the Luria-Bertani (LB) culture medium which was prepared by 10 g tryptone, 5 g yeast extract, and 5 g NaCl dissolving in 1000 mL of purified water (pH7.0-7.2). The LB culture medium was also sterilized by the same abovementioned condition and stored in a refrigerator at 4 °C for next biothermodynamics investigation of E. coli by microcalorimetry.

Sample preparation
About 0.2 g dried S. miltiorrhizae powder (through a 50-mesh sieve) was dissolving by 20 mL 80% MeOH in a 50-mL centrifuge tube and the weight of the tube while sealed up with parafilm was recorded. The tube containing the sample solution was transferred into an ultra-sonic water bath for extraction for 30 min at room temperature. After ultrasonication, the lost weight was made up before following centrifugation at 4000 rpm for 10 min. Next, the supernatant was transferred into a new 30-mL centrifuge tube and stored at 4 °C in the dark until the microcalorimetric measurement.

Microcalorimetric measurement
A 3114/3236 TAM air microcalorimeter (Thermometric AB, Sweden) was used for recording the heat-flow power-time (HFP-t) curves of E. coli growth in the absence (the control) or presence of S. miltiorrhizae solution through ampoule method in batch mode. The microcalorimeter was brought to equilibrium temperature overnight in advance. Four-milliliter LB culture medium containing the E. coli suspensions at the cell density of 1 × 10 6 colony forming units (CFU)/mL was introduced into each 20-mL sterilized ampoule. Then, 1 mL of S. miltiorrhizae solution was added. Correspondingly, the ampoule containing E. coli suspension without the sample solution was set as the control group. Afterwards, the ampoules containing only E. coli suspension and one of the 32 batches of S. miltiorrhizae sample solutions at the final concentration of 10 mg/mL were sealed, shaken-up slightly, and put into the microcalorimeter. Subsequently, after a balance of the instrument for minutes to 37 °C, the HFP-t curves were recorded continuously by Thermometric AB program using the dedicated software package at an interval of 1 min until the curves returned to the baseline. All the experiments were operated in superclean worktable at 37 °C.

Similarity analysis
In reference to the idea and application of similarity analysis (SA) on the chromatographic fingerprints of TCMs  (Zhai et al. 2014;Qin et al. 2015), in this study, SA was introduced for intuitively evaluating the similarities of the changing trends of HFP-t curves of E. coli growth affected by S. miltiorrhizae from different sources based on the original data from these curves. The correlation coefficient for similarity among the HFP-t curves was calculated by cosin method by using the software of Windows SPSS Inc. version 18.0 (Chicago, IL, USA).

Multivariate analysis of variance
As a widely-used statistic method for comparing between-group information, the multivariate analysis of variance (MANOVA) is applied to observe whether there were differences of the antibacterial activities of multiregional S. miltiorrhizae on E. coli. P (probability parameter) <0.05 is regarded as statistically significant. The software of Windows SPSS Inc. version 18.0 was used.

Hierarchical clustering analysis
Hierarchical clustering analysis (HCA) is one of the most commonly used approaches for multivariate analysis, which can classify the objects (samples) into classes (clusters) by means of measuring either the distance or the similarity between the objects. Each object within the same class is similar to the others but different from those in other classes based on a predetermined selection criterion (Zhuang et al. 2011). Heml is an easy-to-use tool with transformation and visualization of multi-dimensional data in a single heat map and can provide a concise but comprehensive presentation of biothermokinetics and multiple clustering strategies for analyzing the data (Deng et al. 2014). Additionally, this software can be recolored, rescaled in a customized manner for visualized evaluation. In this study, the software of Heml (Heat map Illustrator) version 1.0 for Windows (Wuhan, P.R. China) was used for HCA and a method called maximum linkage and the Kendall's tau distance as metric was applied to establish clusters.

Principal component analysis
Too many parameters can be extracted from the heatflow power-time curve of E. coli growth affected by S. miltiorrhizae samples from different regions, which will result in many difficulties for accurately assess the anti-E. coli effects. So, the statistical technology of principal component analysis (PCA) was introduced for next data extraction and analysis. As a multivariate data analysis method, PCA is widely used for searching some underlying factors from multidimensional data that play crucial roles from many confused information. It is an eigenvector-based multivariate analysis tool to explain variance of the multivariate data and further to reduce computation burden and transform the original multivariate data to a smaller and more succinct set of variables, namely principal components (PCs), with orthometric and uncorrelated natures (Zhao et al. 2014). As containing nearly all of the original information, PCs could express the maximal variability of the initial data in a graphical formation as a scores plot which could cluster the sample and further differentiate the samples from different sources according to their antibacterial effects, and the loadings plot allows identification of the main parameters (Wu et al. 2016). Then, the new parameter(s) which are farthest away from the main cluster of variables (parameters) are selected as the crucial indexes for next evaluation.
Here, PCA was operated on mean-normalized data of the nine quantitative parameters from the HFP-t curves of E. coli growth affected by different S. miltiorrhizae samples using software of Simca-P 11.5 (Umetrics AB, Umea, Sweden) (Kong et al. 2010).

Metabolic HFP-t curves of E. coli growth
Performing the microcalorimeter, the normal metabolic thermogenic curve of E. coli growth in the LB culture medium at 37 °C in the absence of any substances was determined in Fig. 1, which presented two stages and four representative phases: the first exponential growth phase (A-B), a stationary phase (B-C), the second exponential growth phase (C-D), and a decline phase (D-E).
Then, the corresponding metabolic thermogenic curves of E. coli affected by S. miltiorrhizae from different sources were recorded and shown in Fig. 2. It could be found that the addition of S. miltiorrhizae solution into the internal system of E. coli growth in the glass ampoule, the metabolism of the bacteria would be influenced and the influences could be intuitively shown from the heights, appearance times of the two peaks in the HFP-t curves. But, these curves were similar and the four phases of the curves were still existed. Compared Fig. 1 Typical HFP-t curve of E. coli growth at 37 °C with the control, the appearance time (t 1 , t 2 ) for first and second highest peaks in the HFP-t curves of E. coli growth affected by S. miltiorrhizae from various sources were prolonged and the highlights decreased, indicating that the sample solutions had inhibitory effects on E. coli growth.

Quantitative thermokinetic parameters for E. coli growth
The HFP-t curves abided by the following equation in the exponential phase of E. coli growth: where P 0 regards as the heat-output power at time t = 0, and P t represents the value of power at anytime during determination. P 1 and P 2 are the heat-flow output powers of the first and second exponential phase of E. coli growth, and t 1 and t 2 are the homologous appearance times. By plotting the logarithm of the HFP-t curve, another important thermokinetic parameter-growth rate constant (k) can be calculated from the slope of the line according to the above-listed equation. There are two exponential growth phases in all HFP-t curves, so the corresponding growth rate constant k 1 is for the first exponential growth phase and k 2 for the second one. Similarly, by integrating the areas under the HFP-t curves, Q 1 is the heat output of the first exponential growth phase, Q 2 is of the second exponential growth phase and Q s is the sum (1) P t = P 0 exp(kt) or lnP t = lnP 0 + kt of the heat output of the whole growth processes. All these thermokinetic parameters including k 1 , k 2 , t 1 , t 2 , P 1 , P 2 , Q 1 , Q 2 and Q s could be read from the HFP-t curves or calculated according to the above-referenced equation, which have been listed in Table 1. The value changes of these thermokinetic parameters could also quantitatively reflect the influence of S. miltiorrhizae on E. coli growth.
However, the bewildering changing trends of the complex quantitative parameters made it virtually difficult to draw a definitive conclusion on the antibacterial activities of S. miltiorrhizae samples from different sources. Therefore, introducing some powerful chemometric methods for simplifying the evaluation was necessary.

SA
It could be found in Fig. 2 that compared with the control, the addition of S. miltiorrhizae sample solutions influenced the growth of E. coli, from the changes of the shape of HFP-t curves and the peak heights and appearance time of the two highest peaks. Then, SA was performed on the values of the nine parameters (k 1 , k 2 , t 1 , t 2 , P 1 , P 2 , Q 1 , Q 2 and Q s ) in Table 1. The results showed that correlation coefficients for similarity among the HFP-t curves of E. coli growth without (the control) and with each S. miltiorrhizae sample solution from different sources were presented as 0.756, 0.871, 0.858, 0.870, Fig. 2 The HFP-t curves of E. coli growth at 37 °C affected by S. miltiorrhizae. Samples from a Sichuan province, b Henan province, c Shandong province, d Beijing city, and e Guizhou province at 10 mg/mL, respectively 0.856, 0.529, 0.868, 0.821, 0.880, 0.794, 0.839, 0.832, 0.827, 0.842, 0.831, 0.831, 0.839 for Guizhou province, 0.858, 0.863, 0.679, 0.872, 0.827, 0.862 for Sichuan province, 0.876, 0.743, 0.772, 0.824 for Shandong province, 0.798, 0.819, 0.744 for Henan province and 0.676, 0.587 for Beijing city. The differences of correlation coefficients illustrated various antibacterial activities of S. miltiorrhizae samples on E. coli growth. Small value of similarity indicated that the HFP-t curve of E. coli growth was significantly influenced by S. miltiorrhizae samples compared with the control. So, it could preliminarily inferred that S. miltiorrhizae samples from Beijing city had the strongest anti-E. coli effects. Nevertheless, it was not sufficient and accurate to evaluate the antibacterial activities only from the results of similarity analysis. So, a further multivariate analysis of variance for between-group information was carried out in the next part.

MANOVA
In order to assess the differences in antibacterial effects of S. miltiorrhizae samples from different provinces or cities, the values of nine parameters including k 1 , k 2 , t 1 , t 2 , P 1 , P 2 , Q 1 , Q 2 and Q s in Table 1 were put into the software of Windows SPSS Inc. version 18.0 for betweengroup multivariate analysis of variance and the values of the Wilks's Lambda, F, and P were obtained in Table 2. From this table, it was apparently found that the Wilks' Lambda value for group of Sichuan province and Beijing city, as well as the group of Shandong province and Beijing city were the smallest with a value of 0.000, followed by the group of Sichuan and Henan provinces (0.001), Sichuan and Shandong provinces (0.002), Shandong and Henan provinces (0.011), Guizhou Province and Beijing City (0.088), Henan Province and Beijing City (0.109), Guizhou and Sichuan provinces (0.114), Guizhou and Henan provinces (0.116), Guizhou and Shandong provinces (0.415). Also, the P values for the group of Guizhou and Shandong provinces, Sichuan and Shandong provinces, Shandong and Henan provinces, as well as Henan province and Beijing city were greater than 0.05, indicating no obvious differences in the anti-E. coli effects of these samples; while, the P values for the group of Guizhou and Sichuan provinces, Guizhou and Henan provinces, Guizhou province and Beijing city, Sichuan and Henan provinces, Sichuan province and Beijing city, Shandong province and Beijing city were smaller than 0.05, showing significant differnces in the anti-E. coli effects of these samples. With the help of MANOVA, the differences of the anti-E. coli effects of the S. miltiorrhizae samples from various provinces or cities could be quickly explored. But, the comparison on the anti-E. coli effects of the samples could not be obtained. So a further chemometric analysis will provide the solution.

HCA
A hierarchical agglomerative cluster analysis was operated based on the nine quantitative parameters in Table 1 to group the tested samples according to their anti-E. coli effects. This analysis found natural groupings of the data set and the heat map of respective samples corresponding to each product area was presented in Fig. 3. The different colors and their changes represented the various degrees of nine parameters for each sample regarding the anti-E. coli effects. It was clear that the 32 samples could be divided into six main clusters, which were colored by six different tincts. Cluster I was consisted of samples XB-1 and XB-2 which were both collected in Guizhou province; Cluster II was composed of samples XB-3, XB-4, XB-5, XB-8, XB-11, XB-13, XB-15, SD-3 and HN-1 samples that were obtained in Guizhou, Shandong and Sichuan provinces, respectively; Cluster III was made up of samples XB-6, HN-2, HN-3, H-1 and T-1 samples from Guizhou Province, Henan Province and Beijing City; Cluster IV included samples XB-7, XB-9, XB-10 and SD-2 samples purchased from Guizhou and Shandong provinces; Cluster V contained samples XB-12, XB-14, XB-16, SD-1 and SD-4 samples delivered from Guizhou and Shandong provinces; Cluster VI was consisted of samples XB-17, FC-1, FC-2, FC-3, FC-4, FC-5 and FC-6 collected in Guizhou and Sichuan provinces. The tested samples were almostly clustered according to their collected sources based on the anti-E. coli effects. But, the samples from Guizhou province were the exceptions, which were segmented and clustered into each group. The possible reasons might be that different harvesting time of the 17 samples led to various contents of the main active constituents, giving rise to the different anti-E. coli effects (He et al. 2010). In addition, from the changes in color, it was difficult to compare the anti-E. coli effects of the 32 samples. The possible reason might be due to the too many parameters, which overlapped or concealed the real information. Therefore, it was important to find the main parameter(s) which could be used for quickly evaluating the anti-E. coli effects of S. miltiorrhizae samples from different sources. Therefore, in the next section, principal component analysis was introduced.

PCA
PCA is a sophisticated technique in applied data analysis work and has a satisfactory ability to simply multivariate variation and confused data set, and only the important or main characteristics of the original data were retained. Here, PCA was operated on mean-normalized data of the nine quantitative parameters including k 1 , k 2 , t 1 , t 2 , P 1 , P 2 , Q 1 , Q 2 and Q s in Table 1. It could be seen from Fig. 4 that the original nine-dimensional space were projected to the new two-dimensional level with the first two PCs (PC1 and PC2) presenting nearly 89% contribution of the original data set. The scores plot in Fig. 4a demonstrated the sources distribution and samples clustering of S. miltiorrhizae. In terms of this scores plot, all samples could be clustered into five groups based on the distribution of each scatter. Group I was consisted of samples XB-1…. XB16 from Guizhou province; group II included samples FC-1, FC-2, FC-3, FC-4, FC-5 and FC-6 from Sichuan province; group III was made up of SD-1, SD-2, SD-3 and SD-4 from Shandong province; group IV was composed of samples HN-1, HN-2 and HN-3 from Henan province; and group V was constituted by samples H-1 and T-1 from Beijing city. The results showed that all tested S. miltiorrhizae samples could be well clustered based on their anti-E. coli effect according to their sources with sample XB-17 as an outlier, which were similar to the above results of HCA. Returning to the curves in Fig. 2e and the data in Table 1, it could be figured out that the second peak of the HFP-t curve for sample XB-17 was the Fig. 4 a Scores and b loadings plots generated from PCA based on the nine quantitative parameters. Possible main parameters were marked with a red dotted circle highest, and the value of P 2 was the biggest, indicating relatively poor anti-E. coli effects of this sample, which might due to its different harvesting time with other sample from Guizhou province (He et al. 2010). The loading plot in Fig. 4b indicated that the parameters k 2 and P 2 were farthest away from the main cluster of the other seven parameters and contributed most for PC1 and PC2, which might be the two underlying quantitative parameters that played crucial roles in evaluating and comparing the antibacterial efficacy of S. miltiorrhizae samples from various sources. Then, the box and whisker plots including the minimum and maximum values (whiskers around the box to indicate the range of the variable), the median (a central point to indicate central tendency), and 25% quartile and 75% quartile values (a box to indicate variability around this central tendency) based on these two parameters, k 2 and P 2 , was performed in Fig. 5 to present the relationships between the values of parameters and the sources of samples. It could be acquired from the plots that the S. miltiorrhizae samples collected from Beijing city exhibited the strongest anti-E. coli activities due to the smallest values of k 2 and P 2 .

Inhibitory ratio (I, %)
In order to quantitatively and intuitionally describe and compare the inhibitory degree of S. Miltiorrhizae samples on E. coli growth, another important and crucial parameter, inhibitory ratio (I, %) was calculated based on the main parameter P 2 according to the following equation: where P (2,0 ) represents the heat-flow output power of E. coli growth in the second exponential phase of the control, P (2,s ) is the heat-flow output power of E. coli growth in the second exponential phase with treatment of S. miltiorrhizae samples. I values have also been listed in Table 1. Some negative values of I indicated that the coresponding samples exhibited relatively poor anti-E. coli effects at the experimental concentration of 10 mg/ mL. Then, for more clear description of I of the samples according to their collected sources, the box and whisker plots, which also included the minimum and maximum value, the median, and 25% quartile and 75% quartile values of I was made in Fig. 6. As could be seen from the plots, the I (%) values for S. Miltiorrhizae samples from Beijing city were the biggest, indicated the strongest anti-E. coli effects, followed by samples from Henan province, Guizhou province, Shandong province and Sichuan province. The result was in accordance to the the above results of HCA and PCA. The above results have indicated antibacterial effects of S. miltiorrhizae on E. coli growth. And the samples from Beijing city might be used as novel and underlying antibacterial candidates for the resistance of E. coli to the existing drugs in the future.

Fig. 5
Box and Whisker plot for the main parameters k 2 and P 2 for S. Miltiorrhizae from different sources on E. coli. This plot was performed using the software of Origin 8.5 showing the minimum and maximum value, the median, 25 and 75% quartile and the range Fig. 6 Box and Whisker plots for the inhibition ratio of S. Miltiorrhizae from different sources on E. coli microcalorimetric method, some important information including the real-time HFP-t curves, as well as some quantitative thermokinetic parameters of E. coli growth affected by S. Miltiorrhizae samples was obtained at the same time, which could not be obtained by traditional microbiological methods (Klančnik et al. 2010;Ahmed et al. 2014). By analyzing the HFP-t curves and the thermokinetic parameters of E. coli growth affected by S. Miltiorrhizae samples using SA, MANOVA, HCA and PCA, it could be quickly found that S. Miltiorrhizae samples exhibited antibacterial effects on E. coli and the samples collected from Beijing city exhibited the strongest anti-E. coli effects, which might be used as novel and underlying antibacterial candidates for the resistance of E. coli to the existing drugs in the future.
This present study also showed that microcalorimetric technology offered some notable advantages for biological investigation compared with some traditional microbiological approaches. This tool could not only save more experiment time, but also exhibit satisfactory sensitivity, accuracy and reproducibility. By the combination of microcalorimetry and chemometrics, the antibacterial effects of other substances including TCMs could be accurately and quickly evaluated, providing a useful method and idea for further study in the light of developing new antibacterial agents with high activity and low toxicity.
In the ongoing study, more bacteria should be selected as the targets to confirm the antibacterial effects of S. Miltiorrhizae samples to provide more evidences for S. Miltiorrhizae as a novel antibacterial agent. In addition, the antibacterial mechanism of S. Miltiorrhizae on E. coli should also be classified for its following application in practice to deal with many essential problems that have not yet been solved due to drug resistance and tolerance after abusing medicines.