Screening of anti-fatigue active ingredients of Eleutherococcus senticosus via spectrum-effect relationship based on factor analysis and LC-MS/MS

Abstract ES contains compounds known to have significant anti-fatigue activity. In recent years, it has received extensive attention because it is efficient. However, its active ingredients on antifatigue effect are still unclear. This study attempts to establish the spectrum-effect relationship of ES antifatigue activity to screen the effective components. The results showed that the similarity of 15 ES fingerprints obtained by LC-MS/MS was 0.533–0.992, and the chemical structures of 22 common peaks were identified. The anti-fatigue activity of 15 batches of ES was characterized by forced swimming test of mice and quantified by CAFI, among which S4, S1 and S5 had better activity. 9 components (caffeic acid, 5-(4-O-β-D-glucosylferoyl)-quinic acid, (±)13-HODE, isofraxidin, eleutheroside E, syringin, pinoresinol diglucoside or its isomer, 7,8-dihydrodehydrocarbinol alcohol-4-O-β-D-glucoside, secoisolariciresinol-4-O-β-D-glucoside) highly related to anti-fatigue activity may be the effective components of ES. Graphical Abstract


Introduction
Eleutherococcus senticosus (Rupr.et Maxim.)Maxim is a shrub mainly distributed in the Far East of Russia, Northeast China, South Korea and Japan, which belongs to Araliaceae (Chen et al. 2021).Its dry root or stem, also known as ES, is listed in the Chinese Pharmacopoeia for long-term use alone or in combination with other TCMs to treat fatigue, insomnia, hyperlipidemia, coronary heart disease (Jia et al. 2021).Among them, the effect of preventing and alleviating sports fatigue is widely recognized (Zhu et al. 2011).Modern pharmacological research shows that the broad-spectrum pharmacological activity of ES is attributed to its multiple secondary metabolic components, such as flavonoids, lignans, polysaccharides, triterpenes and organic acids (Yan et al. 2014).Considering that about 20% of the world's population is in fatigue and the terrible complications caused by long-term fatigue, it is necessary to screen effective anti-fatigue ingredients from ES, which is of great significance for the discovery of plant based anti-fatigue drugs, the chemical synthesis of anti-fatigue agents and quality control of ES (Shen et al. 2021;Zhu et al. 2021a).However, the effective ingredients of ES for anti-fatigue have not yet been clearly clarified.
Faced with the synergism or inhibition between these natural products and the concept of TCM holism, there is still a lack of perfect strategies to screen active ingredients from plant materials.Fortunately, the spectrum-effect relationship proposed by Li et al. (2002) seems to be one of the solutions to this dilemma (Wu et al. 2022).Spectrum-effect relationship explains the relationship between medicinal plant fingerprints and specific functions to find potential effective components, and finally, a quality standard reflecting the internal quality of the product is established, which is basically consistent with the efficacy of the product and its grade (Yu et al. 2022).Spectrum-effect relationship has been successfully applied in some TCMs and gradually aroused people's interest.However, there is a problem worth considering, that is, when the spectrum-effect relationship is used for a specific function of multiple indicator evaluation, it often needs to process a lot of data and may get some irrelevant results, which is time-consuming and cumbersome (Zhang et al. 2022).A comprehensive evaluation based on multiple indicators may be a reasonable answer to the above question.Based on the idea of dimension reduction, FA obtains the comprehensive score through linear weighting of a few comprehensive indicators.By studying the internal correlation of multiple variables, the original variables are grouped according to the correlation, and the indicators with high correlation are expressed by a comprehensive indicator, that is, a common factor.As far as we know, the application of factor analysis in spectrum-effect relationship has not been reported.
In this study, FA was used to obtain the CAFI for the first time.The spectrum-effect relationship between CAFI and fingerprint was constructed by GRA and PLSR to screen potential active components.The purpose of this work is to explore the potential effective components in the anti-fatigue process of ES and the feasibility of factor analysis in the spectrum-effect relationship.

Results of fingerprint
Precision, stability and repeatability tests were carried out to evaluate the robustness of the method.The results showed that RSD were less than 3%, indicating that the analytical method was reliable.
Q-TOF-LC-MS/MS was used to analyze 15 batches of ES, the standard fingerprint and the pictures of 15 batches of ES materials were shown in Figure 1A and 1B, a total of 25 peaks were identified in all samples.The similarity analysis in Table S1 shows that when S1 is used as the reference map, the similarity ranged from 0.533 to 0.992, indicating that the chemical components of ES from different regions were quite different, which was suitable for the study of spectrum-effect relationship.Results of PCA showed that 15 batches of ES couldn't form clusters well, among which S4, S11 and S15 had large differences (Figure 2).
The 25 common peaks of the fingerprint were characterized by LC-MS/MS.The TIC chromatogram in positive ion mode (Figure S1A), negative ion mode (Figure S1B) and HPLC chromatogram (Figure S1C) were used to analysis the structure of the components.Characterization results of common peaks are shown in Table S2.

Evaluation results of anti-fatigue activity of ES
The forced swimming test of mice is one of the most widely used models to evaluate the anti-fatigue performance of subjects (Wang et al. 2021).The anti-fatigue performance of 15 batches of ES was comprehensively evaluated by measuring the exhaustion time and anti-fatigue related biochemical parameters.
The weight change of mice during the experiment is shown in Figure S2A, which shows that no significant difference in weight compared with the control group (p > 0.05), indicating that ES extract had no effect on the weight of mice.Exhaustion  time is the most intuitive evaluation index of the mice forced swimming test (Shen et al. 2021).The effect of ES extract on exhaustion time of mice (Figure S2B) shows that the average exhaustion time of the control group is 256 s, and the exhaustion time range of each administration group is 327 s$520 s.Compared with the control group, all batches ES can significantly prolong the exhaustion time of mice (p < 0.05).Among them, S4 has the best activity with the exhaustion time of 520 s (p < 0.01).The above results show that ES has good anti-fatigue activity, however, the anti-fatigue activity of ES varies greatly in different regions.
The metabolites in the body are not removed and reduced in time, fatigue will occur.BUN and LA are two common metabolites under hypoxia (Zhu et al. 2021b).The changes of BUN and LA are shown in Figure S2C and S2D, the administration group reduced the content of .97%,respectively compared with control group.In terms of reducing BUN content, S12, S13, S14 and S15 samples enhanced therapeutic effects, and S4 and S14 had better effects in reducing LA content.One way ANOVA showed that ES in different regions could significantly reduce the levels of BUN and LA in mice compared with the control group (p < 0.05), and the ability of ES in different regions to reduce BUN and LA was different.
Body movement needs energy supply, Fatigue will occur when energy materials are exhausted (Jia and Zhao 2022).Glycogen is an important energy storage substance in the body, which can be mainly divided into LG and MG.ATP is the direct energy supply material when the body is moving, which releases a lot of energy when it is converted into ADP.ATP deficiency can directly cause muscle contraction disorders.Therefore, the levels of LG, MG and ATP in the body can directly reflect the exercise endurance of the body.The effect of ES on energy storage substances in mice is shown in Figure S2E, S2F and S2G.ES in different regions could significantly increase the LG reserve of mice (p < 0.05).Except for S13 and S14, other samples could significantly increase the MG reserve of mice (p < 0.05).Only S2, S4, S8, S9, S11, S13 and S14 could significantly increase the ATP content of mice (p < 0.05).The results showed that ES could significantly increase the content of energy storage substances in mice, thus providing more endurance for exercise and improving the anti-fatigue ability of mice.
The excessive free radicals produced by oxidative stress react with the membrane structures of cells and organelles, leading to oxidative damage, affecting their metabolic functions, and leading to fatigue (Ma et al. 2022).MDA is the end product of oxidation reaction, which will change the permeability and rheology of cell membrane, and then cause functional damage.Antioxidant capacity of ES is shown in Figure S2H and S2I.Except for the batches of S6, S7, S8, S10 and S12, ES can significantly improve the SOD activity of mice (p < 0.05), and the 15 batches of ES can significantly reduce the production of MDA in mice (p < 0.05).The results showed that ES had a good protective effect on oxidative damage induced by exhaustive exercise in mice.
The effect of ES extract on the main tissue morphology was investigated.Compared with the control group and the quiet group, the volume of liver and gastrocnemius muscle in the administration group increased to varying degrees, with little effect on the volume of heart (Figure S3B), indicating that ES may enhance the antifatigue ability of mice mainly by affecting liver and gastrocnemius muscle.
The myocardium, liver and gastrocnemius muscle tissues of mice in S4, S11 and S15 sample administration groups with large differences from result of PCA were selected for H&E staining.The results are shown in Figure S3A.The myocardial cells of the mice in the quiet group were arranged orderly without cell breakage, and the structure was normal.However, the myocardial cells of control group were arranged disorderly, the blank space area increased, and there was cell breakage, which was improved in the administration group.In the quiet group, the liver tissue morphology and structure were complete, the cells around the central vein were arranged radially closely, the hepatic cords were arranged neatly, the morphology of hepatocytes was normal, and no necrotic cells and other pathological changes were found.In the control group, the hepatocytes were arranged in disorder with no obvious radial shape, the cell volume increased, and the boundary of hepatic cord was not clear.In the administration group, the hepatocyte arrangement was relatively normal, the boundary of hepatic cord was clear, and the cell volume returned to normal.The histopathological section of gastrocnemius muscle showed that the muscle fiber structure of the quiet group was normal, the cell architecture was clear, the nuclei were scattered and arranged orderly, the band structure was neat, and there was no inflammatory cell infiltration.In the control group, the muscle fibers were significantly broken, arranged unevenly, the gap was enlarged, and the cells were edematous.The histopathological results of gastrocnemius muscle in the administration group were significantly improved.It can be seen from the above results that the protective effects of these three batches of AS on mouse tissues are S4 > S15 > S11.
During the forced swimming test of mice, nine anti-fatigue related indexes, that is exhaustion time, BUN, LA, LG, MG, ATP, SOD, MDA were measured to reflect the antifatigue activity of ES from different aspects.FA is proposed to obtain the CAFI representing the anti-fatigue ability of ES.
The variables that establish the FA model are required to have a certain correlation.KMO test and Bartlett sphericity test was used to investigate the correlation between variables.The results showed that KMO ¼ 0.694, p < 0.05, indicating that the variables are related and data structure applicable to FA.
Principal component method was used to extract common factors.The eigen value of principal factors of each common factor is shown in Figure 3A.According to the principle that the eigen value of principal factors >1, two common factors with the cumulative interpretation variance reaches 75.47% are selected.After selecting the first two common factors, the slope slows down, indicating that the common factors increase and have little contribution to the interpretation of the original variables.
After determining the two common factors, in order to make the load of each variable on the two common factors more meaningful, it is necessary to rotate the load of the common factor to differentiate it at the level of 0 and 1.This study adopts the maximum variance rotation method, and the rotation converges after eight iterations.The results are shown in Figure 3B.Common factor 1 mainly explains the information of exhaustion time, LG, MG, and common factor 2 mainly explains the information of BUN, LA, ATP, SOD and MDA.
The common factor, the original variable and the factor score system are linearly combined to calculate the score of each factor (Table S3).According to the score coefficient of each variable on the common factor, the common factor score model is established as Eqs.( 1)-(3).
where f 1 represents the comprehensive score of common factor 1, f 2 represents the comprehensive score of common factor 2, i represents the index of anti-fatigue, i 1 represents the index of exhaustion time, i 2 represents the index of BUN, i 3 represents the index of LA, etc. f represents the comprehensive score of the sample.By comparing the comprehensive score f (Table 1), which is the CAFI, the antifatigue ability of ES is evaluated as S4 According to the PCA result that S4 > S15 > S11, which is consistent with the ranking of its CAFI, indicating that the CAFI obtained by FA can measure the anti-fatigue efficacy of ES.In the literature, it is necessary to establish the spectrum-efficiency relationship between each index and the fingerprint, and finally select the components in the intersection (Zhang et al. 2022).In this work, it is only necessary to establish the spectrum-efficiency relationship between the common peaks of the fingerprint and CAFI, which greatly simplifies the operation.

Screening ES anti-fatigue active components by spectrum-effect relationship
The results of GRA are shown in Table S4.Except for peaks 2 and 20, the correlation degree between other peaks and the CAFI is greater than 0.8, indicating that ES has a significant impact on the anti-fatigue activity, and the anti-fatigue activity of ES is jointly acted by multiple components.The correlation degree of common peaks 1, 3, 4, 5, 6, 7, 8, 9, 10, 11, 13, 14, 15, 16, 21 and 23 is greater than 0.9, indicating that these 16 common peaks are highly correlated with anti-fatigue activity and may be the active components of ES.
PLSR adopts the technology of information synthesis and screening in regression modeling, proposes several comprehensive variables with the best explanatory ability to the system in the independent variable system, and uses them to carry out regression modeling.PLSR is a combination of multiple linear regression analysis, canonical correlation analysis and PCA, which can solve the problem of multiple correlations between variables.The feasibility of PLSR was tested.R2X(cum) represents the cumulative fitting degree of the model to the original data.The closer the value is to 1, the stronger the model interprets the data.Q2(cum) indicates the cumulative prediction ability of the model to the data, and Q2(cum) greater than 0.5 indicates that the model has good prediction ability to the data.It can be seen from Figure 4A that a principal component is selected, R2X(cum) ¼ 0.76, Q2(cum) ¼ 0.49.When two principal components are selected, R2X(cum) ¼ 0.85, Q2(cum) ¼ 0.65.Therefore, the model has a better ability to interpret and predict the data when two principal components are selected.
According to the CoeffCS between CAFI and common peaks (Figure 4B), P4, P17 and P19 are negatively correlated with the CAFI, indicating that the anti-fatigue activity of ES will be reduced when the content of these components increases.All other peaks are positively correlated with the CAFI, indicating that the anti-fatigue activity of ES is enhanced when the content of these components increases.VIP reflects the contribution of each variable to the model, and the threshold is usually set to 1. VIP >1 indicates that it is an important variable in the interpretation of independent variables.The VIP of each peak is shown in Figure 4C.The common peaks with VIP >1 was P9, P6, P23, P15, P13, P7, P11, P16 and P14, indicating that the components represented by these common peaks contribute greatly to the anti-fatigue activity of ES.Based on the analysis results of GRA and PLSR, the common peaks P9, P6, P23, P15, P13, P7, P11, P16 and P14 are speculated to be the active ingredient for the anti-fatigue activity of ES.These common peaks were identified as caffeic acid, 5-(4-O-b-D-glucosylferoyl)-quinic acid, (þ/-)13-HODE, isofraxidin, eleutheroside E, syringin, pinoresinol diglucoside or its isomer, 7,8-dihydrodehydrocarbinol alcohol-4-O-b-D-glucoside, secoisolariciresinol-4-O-b-D-glucoside.The screened ingredients eleutheroside E is consistent with the research results of Huang (Huang et al. 2011).Syringin and (þ)-syringaresinol-D-O-b-D-glucoside have similar chemical structures, and the latter has been shown to have a significant effect on prolonging the exhaustion time of mice (Nishibe et al. 1990).

Materials and reagents
15 batches of dried roots of ES (S1-S15) were collected from China (Table S5), it was identified as ES according to the microscopic identification and TLC described in the Chinese Pharmacopoeia (National Commission of Chinese Pharmacopoeia 2020).Assay kit of LG, MG, ATP, LA, BUN, MDA and SOD were purchased from Nanjing Jiancheng Bioengineering Institute (Nanjing, China).Methanol, acetonitrile (Merck Drugs & Biotechnology, Darmstadt, Germany) and formic acid (Thermo Fisher Scientific, Waltham, MA, USA) were of MS grade.The chromatographic water was prepared by unique multifunctional ultrapure water system (Research Scientific Instruments Co., Xiamen, China).

Establishment of fingerprint of ES
ES materials were crushed and filtered with 40 meshes. 1 g of ES powder fully mixed with 20 mL 70% methanol in a conical flask and sealed.Ultrasonic extraction was carried out at 40 C for 30 min.The final sample solution was obtained by centrifugation (9800Âg, 10 min) and filtrated with 0.22 mm microporous filter, stored at 4 C for subsequent HPLC fingerprint analysis.
The MS detection used an ESI source, ion information was collected in positive and negative ion mode respectively, the ion scanning range was 115-1500 m/z.The dryer temperature was 320 C, the volume flow of dry gas was 8.0 L/min, the sheath gas temperature was 320 C, the capillary voltage was 3.5 kV.The fragment voltage was 110 V, and the secondary voltage was 5, 10, 20, and 30 V. Precursor ions and fragment ions were extracted by MassHunter software (version B.05.00, Agilent Technologies) for chemical structure determination.
S4 was randomly selected as the experimental object and prepared sample solutions.The precision experiment was carried out by continuously injected S4 sample solution for 6 times.The stability experiment was validated by injected S4 sample solution at 0, 2, 4, 8, 12 and 24 h.The repeatability experiment was performed by detected 6 S4 sample solutions prepared in parallel.The common peak area and retention time was recorded, and RSD was calculated to evaluate the feasibility of the method.
The similarity evaluation system of traditional Chinese medicine fingerprints (Version, 2012) was used to calculate similarity, and principal component analysis (PCA) was implemented on SIMCA 13.0 software (Umetrics AB, Umea, Sweden).

Determination of anti-fatigue activity of ES
A total of 102 SPF KM mice were purchased from Liaoning Changsheng Biotechnology Co., Ltd (Benxi, China), female, weighing 18 6 2 g.They were kept in a ventilated environment with 25 6 1 C, 50 6 10% humidity, 12 h of light, food and water were unlimited.The European Community guidelines for the use of laboratory animals were used to regulate the operation on experimental mice.The ethical approval number of animal experiment is 2022072.
An appropriate amount of ES powder and 70% methanol were mixed in the ratio of 1:20 (g: mL) of material and solvent, sealed for ultrasonic extraction for 30 min at room temperature, filtered and volatilized the solvent, added with normal saline for dissolution, stored at 4 C for use.After 5 days of normal feeding, the mice were divided into administration group (S1-S15), control group (Control) and quiet group (Blank) by random number table, with 6 mice in each group.The administration group was orally given a dose of 400 mg/kg body weight, and the control group and quiet group were orally given normal saline, with a volume of 0.1 mL/10 g body weight at 9:00 a.m.every day, lasted for 28 days.The exhaustion test was conducted at a 51 cm Â 29 cm Â 38 cm (length, width, height) water tank, water temperature 25 6 2 C, water depth 30 cm. 30 min after the last administration, each mouse was tied with a 10% body weight lead sheath at the tail and put into the water to force the mouse to swim.When the mouse was floating, the water surrounding was stirred to make its limbs paddle and swim.When the head of mice was completely immersed in the water for more than 10 s as exhausted, and taken it out immediately.The exhaustion time of the mice was recorded as the time when the mice were put into the water until exhausted.After the forced swimming test of mice, the eyeballs of the mice were removed to take blood.The blood was coagulated at room temperature and centrifuged (9800Âg, 10 min).The supernatant was used to detect the contents of LA and BUN.The mice were killed immediately after blood collection.The myocardium, liver and gastrocnemius muscle of mice were isolated, and the LG, MG, ATP, MDA and SOD in liver or muscle tissue were measured according to kit usage recommendations.An appropriate amount of tissue was taken to prepare histopathological sections.The changes of cell structure were observed under the microscope after H&E staining.The changes of the body weight of mice were monitored at the beginning and end of the forced swimming test of mice.
Negative indicators BUN, LA and MDA were transformed into positive indicators as Eq. ( 4).
where X 0 was the conversion value of index X, and M was the maximum value of index.
The original variables (X 1 , X 2 , Á Á Á , X n ) are first divided into p groups according to their correlations, and the information contained in original variables are replaced by unrelated common factors (f 1 , f 2 , Á Á Á , f p ).The score of each common factor is established as f 1 ¼ a 1 X 1 þ a 2 X 2 þ . . .þ a n X n (a n is the load of X n on f 1 ), the comprehensive scoring model is described as is the variance interpretation of f p ).

Chemometric analysis
FA was performed on SPSS 20.0 (IBM, Chicago, IL, USA) software to obtain CAFI, which was regarded as the parent sequence, and each common peak was a subsequence, GRA was implemented on the SPSSPRO (https://www.spsspro.com/analysis/index),and PLSR was carried out on SIMCA 13.0 (Umetrics, Umea, Sweden).

Statistical analysis
All data were expressed as mean ± SD.Data analysis and visualization were completed on GraphPad Prism 8.0 (GraphPad Software Inc., San Diego, CA, USA).One-way ANOVA test was used to compare the differences between groups.p < 0.05 was considered to be significant.

Conclusion
In this study, the possible anti-fatigue active components of ES were screened through spectrum-effect relationship, they are caffeic acid, 5-(4-O-b-D-glucosylferoyl)-quinic acid, (þ/-)13-HODE, isofraxidin, eleutheroside E, syringin, pinoresinol diglucoside or its isomer, 7,8-dihydrodehydrocarbinol alcohol-4-O-b-D-glucoside and secoisolariciresinol-4-O-b-D-glucoside.It is the first time to try to deal with the spectrum-effect relationship through FA.The method developed in this work is time-saving, simple and efficient.The research results are conducive to improving the level of ES quality control system using spectrum-effect relationship based on FA and LC-MS/MS.

Figure 1 .
Figure 1.HPLC standard fingerprint of 15 batches of ES (A) and pictures of ES (B).

Figure 2 .
Figure 2. PCA of ES based on 25 common peaks.

Figure 3 .
Figure 3. Scree plot of FA (A), Composition diagram in rotating space of FA (B).

Table 1 .
Common factor scores and comprehensive score of ES for anti-fatigue.