Abstract
Steamed Panax notoginseng (SPN) has been used as a tonic to improve the
blood deficiency syndrome (BDS) in the theory of traditional Chinese
medicine. Here, we aim to unveil active constituents and potential
targets related to the hematinic effect of SPN, which has not been
answered before. In the study a constituent-target-disease network was
constructed by combining the SPN-specific and anemia-specific target
proteins with protein-protein interactions. And the network
pharmacology was used to screen out the underlying targets and
mechanisms of SPN treating anemia. Also, the multivariate data analyses
were performed for the double screening. According to the results, 11
targets related to chemical constituents of SPN were found to be
closely associated with the hematinic effect of SPN. Among them, the
direct target protein of mitochondrial ferrochelatase (FECH) had the
major role through the metabolic pathway. Meanwhile, Rk[3] and
20(S)-Rg[3] were predicted to be major constituents related to the
hematinic effect of SPN by both multivariate data analyses and network
pharmacology. And it was been validated by the pharmacologic tests that
Rk[3] and 20(S)-Rg[3] could significantly increase the levels of blood
routine parameters, FECH and its downstream protein of heme in mice
with BDS. The study provides evidences for the mechanism understanding
and drug development of SPN for the treatment of anemia.
Keywords: steamed Panax notoginseng, hematinic effect, active
constituents, mechanism, network pharmacology, multivariate data
analyses
Introduction
Blood deficiency syndrome (BDS) is a common syndrome with high
incidence in clinic of traditional Chinese medicine (TCM), which often
occurs in patients with anemia. It is characterized by a decreased
quantity of red blood cells (RBC) and white blood cells (WBC), usually
accompanied by the diminished hemoglobin (Hb) levels or altered RBC
morphology ([35]Kassebaum et al., 2014). Patients with BDS are often
accompanied with tachycardia, dizziness, shortness of breath, poor
ability to exercise, and even loss of consciousness ([36]Janz et al.,
2013). The causes giving rise to BDS are similar to anemia, including
the reduced erythropoiesis, excessive hemorrhage, and destruction of
RBC, of which the last one mainly refers to hemolytic anemia ([37]Tian
et al., 2017). Hemolytic anemia can be induced by the administration of
acetylphenylhydrazine (APH) and cyclophosphamide (CTX), which is also a
classical model for BDS ([38]Zhang et al., 2014). Today, hemolytic
anemia is mainly treated with blood transfusion. This treatment is
associated with mechanical shearing forces that accelerate red blood
cell rupture, and lead to severe clinical complications, including
intravascular hemolysis, tissue oxidative stress, and multi-organ
dysfunctions ([39]Baek et al., 2012). Therefore, the development of
effective therapies or drugs with blood-enriching efficacy is
beneficial for the prevention and treatment of the disease.
Panax notoginseng (PN) (Burk.) F. H. Chen, a well known medicinal herb
in Asia, has been used to treat blood disorders for thousands of years
([40]Li et al., 2017). In 1944, PN and the relative products were
classified as dietary supplements according to the United States
Dietary Supplement Health and Education Act ([41]103rd Congress, 1994).
Traditionally, PN has been used in both non-steamed and steamed forms.
Unlike the non-steamed one treating bleeding and removing blood stasis,
the steamed PN (SPN) is used as a tonic to enrich blood and tonify the
body, which can improve the BDS by increasing the production of various
blood cells in anemic conditions ([42]Ge et al., 2015; [43]Gu et al.,
2015). In our previous study ([44]Xiong et al., 2017b), we found that
the treatment of SPN could reverse significantly the decrease of levels
of WBC, RBC, Hb, and platelet (PLT) of anemic mice induced by APH and
CTX, which were inapparent when treated with non-steamed PN. The result
was consistent with the report from [45]Zhou et al. (2014). This could
be due to the variation in the chemical composition of PN during the
steaming process ([46]Wang et al., 2012a). Despite various studies on
the processing methods, chemical components and bioeffects of PN
([47]Lau et al., 2009; [48]Sun et al., 2010), much less attention has
been paid on its steamed form, let alone the specific active
constituents and the underlying mechanisms related to the hematinic
effect of SPN, which hinders the development and application of this
valuable medicine.
Network pharmacology, as a system biology-based methodology, offers an
effective approach to evaluate the pharmacological effects of herbal
medicines at the molecular level by predicting the complex interactions
of small molecules and proteins in a biological system ([49]Tang et
al., 2015; [50]Li et al., 2017). It is considered to be a promising way
to unveil properties of herbal medicines and provide valuable insights
into current drug discovery and development. Compared with conventional
“one target, one drug, one disease” mode, network pharmacology focuses
on “multi-targets, multi-constituents treatment to diseases,” which
coincides with the holistic and systematic concepts of TCM ([51]Li et
al., 2015; [52]Quan et al., 2016). It was reported that network
pharmacology has been applied to detect new pharmacologic effects of
herbal medicines, to uncover the interactions between herbal
compounds/formulas and complex syndrome systems, to determine the
active constituents and their mechanisms of action ([53]Tao et al.,
2013; [54]Sheng et al., 2014). Therefore, to better understand the
molecular basis of the enriching-blood effect of SPN, we
computationally recognized the active constituents and potential
targets of SPN treating anemia by the network pharmacology approach
coupled with multivariate data analyses, and experimentally validated
the predicted results.
Materials and Methods
Computational Prediction of Hematinic Constituents and Targets of SPN Using
Network Pharmacology Analyses
Database Construction
Based on the literature research and our previous works on chemical
analysis of SPN ([55]Sun et al., 2010; [56]Xiong et al., 2017a,[57]b),
20 compounds were selected, including ginsenosides of F[2] (1), Rb[1]
(2), Rb[2] (3), Rb[3] (4), Rc (5), Rd (6), Re (7), Rg[1] (8), Rh[2]
(9), Rh[4] (10), Rk[3] (11), 20(R)-Rg[2] (12), 20(S)-Rg[2] (13),
20(R)-Rg[3] (14), 20(S)-Rg[3] (15), 20(R)-Rh[1] (16), and 20(S)-Rh[1]
(17); and notoginsenosides of C (18), R[1] (19), and R[2] (20) in SPN.
The chemical structures of the composite compounds in SPN were obtained
from TCM Database@Taiwan (TDT) or drawn with ChemDraw professional 15.0
([58]Chen, 2011). The targets of constituents were predicted by the
online target prediction software of PharmMapper with a criterion of
“fit score” >4^[59]1 ([60]Wang et al., 2017). Gene and protein targets
associated with the disease of anemia were collected from the Online
Mendelian Inheritance in Man (OMIM) database ([61]Amberger et al.,
2015). Database of Interacting Proteins for protein-protein
interactions (PPI) was employed to identify the possible interactions
of the aforementioned targets. And all protein ID codes were converted
to UniProt IDs ([62]Wei et al., 2016).
Network Construction and Analysis
To provide the scientific and reasonable interpretation of the complex
relationships between the constituents and targets associated with
anemia, network analysis was performed. The chemical constituents, SPN
putative targets, and anemia targets were all connected to construct a
“constituent-target-disease” network with PPI information. Cytoscape
4.3 ([63]Smoot et al., 2011) was applied to visualize and analyze the
network, and calculate the topological features of each node in the
network. Only the hub nodes (twofolds above the median “degree” value
of all nodes) with higher values of “betweenness centrality” and
“closeness centrality” (above the median value of all nodes) were
identified as the candidate SPN targets for anemia.
Targets and Pathways Analyses
To unveil the mechanism of SPN treatment of anemia, DAVID Functional
Annotation Bioinformatics Microarray Analysis^[64]2 was performed
([65]Dennis et al., 2003) for the pathway enrichment analysis. The key
target in the most significant enriched pathway was verified by
performing in vivo experiment in the SPN treatment on anemia.
Screening Hematinic Constituents of SPN Based on the Fingerprint-Effect
Analyses
Chemicals
The reference standards of ginsenosides 20 (S)-Rg[3] and Rk[3] with a
purity ≥ 98% were purchased from the National Institutes for the
Control of Pharmaceutical and Biological Products (Beijing, China).
Methyl alcohol and acetonitrile of HPLC grade were purchased from
Sigma-Aldrich, Inc. (St. Louis, MO, United States). Ultrapure water was
generated with an UPT-I-20T ultrapure water system (Chengdu Ultrapure
Technology, Inc., Chengdu, Sichuan, China). APH was purchased from
HuaXia Chemical Reagent Co., Ltd. (Chengdu, China). CTX was purchased
from Xiya Chemical Industry Co., Ltd. (Shangdong, China). Mouse
ferrochelatase (FECH) enzyme-linked immunoassay kit and heme
enzyme-linked immunosorbent assay kit were purchased from Shanghai
MLBIO Biotechnology Co., Ltd. (Shanghai, China). All other chemicals
used were of analytical grade.
Sample Preparation
The preparation of SPN refers to our previous study ([66]Xiong et al.,
2017a). “Samples were obtained from a single batch of PN root in
Yunnan, China. Steamed PN samples were prepared by steaming the crushed
raw PN in an autoclave (Shanghai, China) for 2, 4, 6, 8, and 10 h at
105, 110, and 120°C, respectively. The steamed powder was then dried in
a heating-air drying oven at about 45°C to constant weight, then
powdered and sieved through a 40-mesh sieve.”
Animals
Animal experimental procedures in the study were strictly conformed to
the Guide for the Care and Use of Laboratory Animals and related ethics
regulations of Kunming University of Science and Technology. The
protocol was approved by the Experimental Animal Welfare and Ethics
Committee, Kunming University of Science and Technology. The
experimental method refers to our previous study ([67]Xiong et al.,
2017a), that “Kunming mice, male and female, weighing 18–22 g, were
purchased from TianQin Biotechnology Co., Ltd., Changsha, Hunan [SCXK
(Xiang) 2014-0011]. Before the experiments, the mice were given
one-week acclimation period in a laboratory at room temperature
(20–25°C) and constant humidity (40–70%), and fed with standard rodent
chow and tap water freely.”
HPLC Analyses
The sample solutions were prepared according to the method in our
previous research ([68]Xiong et al., 2017a). “HPLC analyses were done
on an Agilent 1260 series system (Agilent Technologies, Santa Clara,
CA, United States) consisting of a G1311B Pump, a G4212B diode array
detector, and a G1329B autosampler. A Vision HT C[18] column (250 mm ×
4.6 mm, 5 μm) was adopted for the analyses. The mobile phase consisted
of A (ultra pure water) and B (acetonitrile). The gradient mode was as
follows: 0–20 min, 80% A; 20–45 min, 54% A; 45–55 min, 45% A; 55–60
min, 45% A; 60–65 min, 100% B; 65–70 min, 80% A; 70–90 min, 80% A. The
flow rate was set at 1.0 ml/min. The detection wavelength was set at
203 nm. The column temperature was set at 30°C and sample volume was
set at 10 μl.”
Blood Routine Test
210 km mice, half male and half female, were randomly divided into
seven groups, namely the control group, model group, Fufang E’jiao
Jiang (FEJ) group, and drug groups including raw PN (S1-S3), SPN at
105°C (S4-S8), SPN at 110°C (S9-S13), and SPN at 120°C (S14-S18), 10
mice in each group. The APH and CTX-induced anemia model was applied to
evaluate the “blood enriching” function of PN combined with previous
methods ([69]He et al., 2015). The anemia model was established by
intraperitoneal injected of CTX of 0.07 g/kg for the first 3 days and
hypodermic injection of APH of 0.02 g/kg at the fourth day. Mice in the
control group were administered with 0.9% normal saline, whereas other
groups were administered with FEJ (8 ml/kg), and SPN samples at
different steamed conditions (0.9 g/kg), respectively, by gavage for 12
days. Then the blood was collected for the routing analysis, including
levels of WBC, RBC, Hb, and PLT after 30 min of the last
administration. And the liver tissues were collected for the
determination of FECH and heme levels.
Multivariate Data Analyses
Canonical correlation analysis (CCA)
Canonical correlation analysis is a multivariate analysis used to study
the correlation between two sets of variables. It is used for the
dimension reduction of PCA and to extract the main principal
components, and then describes the whole linear relationship of two
sets of variables by the relevance of two principal components ([70]Shi
et al., 2016). In our study, CCA was used to analyze the relevance
between the peak area values from the HPLC fingerprints and blood
parameters data.
Partial least squares regression (PLSR)
Partial least squares regression is performed to find the inner
relationship between the independent variables (X) and dependent
variables (Y), which are simultaneously modeled by taking into account
X variance, and the covariance between X and Y ([71]Martens and Naes,
1991). In our study, the X matrix is composed of the enhanced
fingerprints, and the Y vector is constructed with the reference values
of hematinic effect obtained by measuring the levels of WBC, RBC, Hb,
and PLT. Then, X and Y are decomposed in a product of another two
matrices of scores and loadings; as described by the following
equations:
[MATH:
X = TPT+E :MATH]
(1)
[MATH:
Y = UQT+F :MATH]
(2)
Where TP^T approximates to the chromatographic data and UQ^T to the
true Y values; notice that the relationship between T and U scores is a
summary of the relationship between X and Y. The terms E and F from the
equations are error matrices. Hence, the PLS algorithm attempts to find
latent variables that maximize the amount of variation explained in X
that is relevant for predicting Y; i.e., capture variance and achieve
correlation ([72]Sundberg, 2008).
Experimental Validation for the Predicted Results
Validation for the Screened Active Constituents
Peaks in the HPLC profile predicted to be responsible for the hematinic
activity of SPN were then identified by reference standards, of which
the activities were finally verified by pharmacologic evaluation using
the methods described in the section of “Blood routine test.” 90 km
mice, half male and half female, were randomly divided into seven
groups, namely the control group, model group, FEJ group, and drug
groups [including low, moderate, and high-dose ginsenoside 20(S)-Rg[3]
group; and low, moderate, and high-dose ginsenoside Rk[3] group], 10
mice in each group. Mice in the control group were intraperitoneal
injected with 0.9% normal saline, whereas other groups were
intraperitoneal injected with FEJ (8 ml/kg), 20(S)-Rg[3] (2.5, 5, and
10 mg/kg, respectively), and Rk[3] (2.5, 5, and 10 mg/kg,
respectively), respectively.
Validation for the Predicted Targets and Its Downstream Protein
The livers of mice from different groups were removed to detect the
levels of FECH and heme. The livers of different groups of mice were
homogenized with a homogenizer and centrifuged for 20 min to collect
the supernatant. Mouse FECH enzyme-linked immunosorbent assay kit and
mouse heme enzyme-linked immunosorbent assay kit were used to detect
the levels of FECH and heme in the supernatant.
Statistical Analyses
All data were expressed as means ± SD. SPSS 21.0 software (Statistical
Program for Social Sciences, SPSS Inc., United States) was applied to
carry out the two-tailed unpaired t-test. Umetrics SIMCA-P 11.5
software (Sartorius Stedim Biotech, Sweden) was applied for PLSR
analysis. CCA was performed using MATLAB 7.0 (Matrix Laboratory, United
States). A value of P < 0.05 was considered to be significant
difference. A value of P < 0.01 was considered to be highly significant
difference. EC[50] value was fitted by probit regression with Origin
7.5 software for windows (OriginLab Corporation, United States).
Results
Computational Prediction Using Network Pharmacology Analyses
Constituents and Targets Prediction
On the basis of database construction, 203 putative targets with “fit
score” >4 were predicted by PharmMapper for 14 compounds and 90
candidate protein targets associated with anemia therapy were collected
by keyword-based searching over the OMIM database, which included 19
common targets out of SPN constituents and anemia. Therefore,
ginsenosides 20 (R)-Rg[2], 20 (S)-Rg[2], Rb[3], Rb[1], F[2], and Rc
were eliminated due to their low binding affinity to all the candidate
targets.
Network Construction
The “constituent-target-disease” network was constructed and the
noteworthy features of the network analyzed could provide some
important information for us to understand the “drug–target”
interaction mechanism of a certain drug on a specific disease. Our
study was focused on the effect of SPN on treating anemia. In Figure
[73]1, the network for the constituents and their potential targets was
illustrated with color-coded nodes. The intermolecular interactions
(constituent-target or target- disease interactions) were indicated as
links, i.e., edges between nodes ([74]Tao et al., 2013). The red
triangles represented the analyzed constituents of SPN, the blue dots
represented the indirect targets of those constituents, the yellow dots
represented the targets of anemia, the purple dots represented the
interactional proteins of the anemia targets and SPN constituents, and
the yellow squares represented the common targets of SPN constituents
and anemia. Obviously, the common targets, as the directed targets of
SPN on treating anemia, were relatively important for further
screening.
FIGURE 1.
FIGURE 1
[75]Open in a new tab
The “constituent–target–disease” network for the treatment of anemia
with SPN. The red triangles represented the analyzed constituents of
SPN, the blue dots represented the indirect targets of those
constituents, the yellow dots represented the targets of anemia, the
purple dots represented the interactional proteins of the anemia
targets and SPN constituents, and the yellow squares represented the
common targets of SPN constituents and anemia. SPN, steamed Panax
notoginseng.
Based on the network analysis, three topological parameters of
“degree,” “betweenness centrality,” and “closeness centrality” were
chosen to screen the potential anemia targets that SPN might affect.
After calculating the values of the three parameters for each
significant protein in the PPI network, the median values of “degree,”
“betweenness centrality,” and “closeness centrality” were 1, 0, and
0.2183, respectively. The protein targets of which the “degree” was
more than twofolds of the median value, and “betweenness centrality”
and “closeness centrality” were higher than the median value, were
chosen as the major targets of SPN treating anemia ([76]Wang et al.,
2018). As shown in Table [77]1, we finally determined that 11 common
protein targets with degree ≧2, betweenness centrality ≧0, and
closeness centrality ≧0.2183 for anemia therapy.
Table 1.
The information of common target proteins and their corresponding
active constituents predicted by network pharmacology analyses.
UniProt No. Protein name Closeness centrality Degree Betweenness
centrality Constituent
[78]P22830 FECH, mitochondrial 0.2810 9 0.0011 (10) (11) (14) (15) (16)
(17) (20)
[79]P16442 Histo-blood group ABO system transferase 0.2995 16 0.0041
(3) (6) (7) (9) (10) (11) (14) (15) (17) (18) (19) (20)
[80]P06702 Protein S100-A9 0.2811 8 0.0036 (3) (6) (7) (8) (10) (11)
(17) (19)
[81]P60568 Interleukin-2 0.2404 4 0.0056 (3) (14)
[82]P06744 Glucose-6-phosphate isomerase 0.2490 2 0.0001 (20)
[83]P30613 Pyruvate kinase PKLR 0.2747 4 0.0014 (7) (20)
[84]Q06124 Tyrosine-protein phosphatase non-receptor type 11 0.2812 6
0.0078 (6) (10) (11)
[85]P19367 Hexokinase-1 0.2635 2 0.0005 (7)
[86]P00374 Dihydrofolate reductase 0.2677 2 0.0006 (7)
[87]P35228 Nitric oxide synthase, inducible 0.2552 4 0.0114 (7)
[88]P42224 Signal transducer and activator of transcription
1-alpha/beta 0.2188 2 0.0024 (18)
[89]Open in a new tab
Pathway Analysis of SPN Treating Anemia
Kyoto Encyclopedia of Genes and Genomes (KEGG) is a knowledge base for
systematic analysis of gene functions in terms of the networks of genes
and molecules. The major component of KEGG is the pathway database that
consists of graphical diagrams of biochemical pathways including most
of the known metabolic pathways and some of the known regulatory
pathways ([90]Kanehisa and Goto, 2000). As shown in Figure [91]2, 10
KEGG pathways were enriched by the pathway-enrichment analysis. And
there was the maximum quantity of targets involved in the metabolic
pathways. Based on the analysis of KEGG, these target proteins were
[92]P22830 (FECH, mitochondrial), [93]P16442 (histo-blood group ABO
system transferase), [94]P00374 (dihydrofolate, reductase), [95]P06744
(glucose-6-phosphate isomerase), [96]P19367 (hexokinase-1), [97]P35228
(nitric oxide synthase, inducible), and [98]P30613 (pyruvate kinase
PKLR). Combined with the results in Table [99]1, FECH and histo-blood
group ABO system transferase showed higher values of degree than
others. And based on the literature research, FECH was reported to be a
mitochondrial membrane-associated protein catalyzing the terminal step
of heme biosynthesis ([100]Yoon and Cowan, 2004), of which the abnormal
synthesis can lead to anemia ([101]Iolascon et al., 2009). Therefore,
the protein of FECH and its downstream protein of heme (Figure [102]3)
were chosen from the map of metabolic pathways for further
verification. Constituents related to the target of FECH included
ginsenosides Rh[4] (10), Rk[3] (11), 20(R)-Rg[3] (14), 20(S)-Rg[3]
(15), 20(R)-Rh[1] (16), 20(S)-Rh[1] (17), and notoginsenoside R[2]
(20).
FIGURE 2.
FIGURE 2
[103]Open in a new tab
Ten enriched pathways with putative targets. X-axis represents the
percentage of targets involved in the pathway. Y-axis represents the
name of putative pathway.
FIGURE 3.
FIGURE 3
[104]Open in a new tab
The structures of (A) FECH and (B) heme. FECH, ferrochelatase.
Prediction of Hematinic Constituents of SPN Based on Multivariate Data
Analyses
HPLC Analyses
HPLC fingerprints for 18 batches of PN samples were shown in Figure
[105]4 ([106]Xiong et al., 2017a). “Peaks with good segregation, which
also occupied large areas from consecutive peaks, were determined as
the common peaks of PN samples. Therefore, fifteen peaks were selected
by comparing their ultraviolet spectra and HPLC retention time.” The
areas of 15 peaks in 18 batches of PN samples were listed in
Supplementary Table [107]S1. “The peak area was defined as 0 for peaks
lacked in chromatograms. The coefficients of variance for almost all
common peaks were higher than 46.6%. This is due to the diversity in
the levels of constituents contained in samples under different process
conditions. The areas of 15 common peaks were used for the following
analysis” ([108]Xiong et al., 2017a).
FIGURE 4.
FIGURE 4
[109]Open in a new tab
HPLC fingerprints of 18 batches of PN extracts. HPLC analyses were done
on a Vision HT C[18] column (250 mm × 4.6 mm, 5 μm) at 30°C. The mobile
phase consisting of A (ultra pure water) and B (acetonitrile) was used
at a flow rate of 1.0 ml/min as the following gradient mode: 0–20 min,
80% A; 20–45 min, 54% A; 45–55 min, 45% A; 55–60 min, 45% A; 60–65 min,
100% B; 65–70 min, 80% A; and 70–90 min, 80% A. The detection
wavelength was set at 203 nm and the injection column was set at 10 μl.
PN, Panax notoginseng ([110]Xiong et al., 2017a).
Blood Routine Test
After the administration for 15 days, the quantities of WBC, RBC, Hb
and PLT from the peripheral blood of mice were shown in Figure [111]5.
Compared with the control group, the levels of WBC, RBC, Hb, and PLT in
the model group were significantly decreased (P < 0.01), indicating the
anemia model was successfully established. Compared with the model
group, WBC, RBC, Hb, and PLT levels in the FEJ and all of PN groups
were increased at different degrees. Besides, there were more
significant differences in the levels of the above four parameters
between the model group and SPN groups steamed at higher temperature
and longer time, suggesting that SPN steamed at higher temperature and
longer time could significantly reverse the decrease of the quantities
of WBC, RBC, Hb, and PLT. While for mice treated with raw PN, the level
of RBC was significantly increased, whereas there was no significant
difference in levels of WBC, Hb, and PLT between raw PN and the model
group, indicating that the blood-enriching effect of raw PN was
generally weaker than SPN. According the results, SPN was observed to
enhance the hematopoietic effect on mice with chemotherapy-induced
anemia, which was consistent with the traditional use of SPN ([112]Gu
et al., 2015).
FIGURE 5.
FIGURE 5
[113]Open in a new tab
The contents of (A) WBC, (B) RBC, (C) Hb, and (D) PLT in the blood of
mice after treated with different PN samples. Each value represents
means ± SD (n = 10). ^∗∗P < 0.01, compared with the control group; ^△ P
< 0.05 and ^△ △ P < 0.01, compared with the model group. FEJ, Fufang
E’jiao Jiang; Hb, hemoglobin; PLT, platelet; PN, Panax notoginseng;
RBC, red blood cell; SD, standard deviation; WBC, white blood cell.
Uncovering Active Constituents by Multivariate Data Analyses
CCA
Canonical correlation analysis was used to establish the
fingerprint-effect relationships between area values of 15 common peaks
in the HPLC data and four blood routine parameters (WBC, RBC, Hb, and
PLT). The analysis result was shown in Table [114]2. The correlation
coefficients showed that the four parameters were positively correlated
with X[4], X[5], X[9], X[10], X[11], X[12], X[13], X[14], and X[15].
Besides, eight peaks: X[5], X[9], X[10], X[11], X[12], X[13], X[14],
and X[15] were highly correlated (| R| > 0.6) with the blood
parameters. This indicates that the decrease of the quantities of WBC,
RBC, Hb, and PLT might be reversed by these compounds.
Table 2.
The correlation coefficients between the common characteristic peaks
and four blood parameters.
Blood parameters Common characteristic peaks
__________________________________________________________________
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
WBC -0.689 -0.789 -0.595 0.122 0.5932 -0.590 -0.662 -0.274 0.740 0.782
0.783 0.805 0.738 0.684 0.747
RBC -0.702 -0.777 -0.601 0.338 0.7276 -0.442 -0.611 -0.181 0.798 0.785
0.782 0.763 0.753 0.644 0.744
Hb -0.854 -0.845 -0.712 0.156 0.757 -0.653 -0.748 -0.170 0.824 0.867
0.862 0.845 0.790 0.671 0.788
PLT -0.845 -0.934 -0.776 0.212 0.7195 -0.676 -0.781 -0.479 0.905 0.940
0.939 0.909 0.907 0.886 0.933
[115]Open in a new tab
PLSR
The PLSR models to correlate chromatographic data and hematinic effect
of 18 batches of PN samples were constructed. Since the total number of
samples (18) was small and the prediction for new samples was not our
first concern, no division was made into a calibration set to build a
PLSR model and a test set to validate the predictive properties. Our
main concern was to focus on the indication of hematinic effect peaks
from the modeling results. PLSR models were built from the normalized
data matrix X containing the 18 PN fingerprints and the response matrix
Y (including Y[1], Y[2], Y[3], and Y[4], which represented WBC, RBC,
Hb, and PLT, respectively). For the model, four principle components
were achieved, accounting for an explained variance of 86.4% for X
variable, 87.1% for Y variable, and a predictive ability (Q^2) of 81.5%
(Supplementary Table [116]S2), indicating that the obtained model was
excellent. As shown in the regression coefficients plot (Figure
[117]6), peaks 1, 3, 6, and 8–15 were positively correlated with the
quantities of WBC, RBC, Hb, and PLT, whereas peaks 2, 4, 5, and 7 were
negatively correlated with the quantities of the four parameters.
FIGURE 6.
FIGURE 6
[118]Open in a new tab
Regression coefficient plots of hematinic effect of 15 peaks in the
chromatograms of PN samples. PN, Panax notoginseng.
Besides, the importance of the X-variables for the model could be
summarized by variable importance for the projection (VIP) values
(usually with a threshold >1.0). Thus, constituents corresponding to
peaks 5, 9, 11, 12, 14, and 15, of which the VIP values were >1.0
(Table [119]3) with high absolute values of coefficients were
considered to be highly related to the hematinic effect of different PN
samples. Furthermore, false discovery rate (FDR, usually with a
threshold ≤0.05) can effectively solve the control of false positive
error in multiple comparisons of high-dimensional data, and can
significantly improve the efficiency of hypothesis testing
([120]Benjamini and Hochberg, 1995). Therefore, constituents
corresponding to peaks 5, 7, 9, 10, and 12, of which the FDR values
were ≤ 0.05, indicated that constituents corresponding to peaks 5, 7,
9, 10, and 12 were positively correlated with the hematinic effect.
Among them, only peak 10 and 12 had significant correlation with the
hematinic effect by P-value correcting (P < 0.05).
Table 3.
Variable importance for the projection of PLSR.
Peak 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
VIP 0.987 0.70 0.56 0.29 1.03 0.37 0.57 0.45 1.121 1.37 1.24 1.41 0.902
1.09 1.23
FDR 0.097 0.085 0.092 0.088 0.041 0.067 0.034 0.068 0.042 0.048 0.072
0.022 0.099 0.061 0.060
P-value 0.091 0.097 0.098 0.558 0.055 0.150 0.0.235 0.185 0.235 0.079
0.011 0.062 0.001 0.086 0.079 0.058
[121]Open in a new tab
Identification of active constituents corresponding to predicted peaks
Based on CCA and PLSR results, constituents corresponding to peaks 10
and 12 were predicted to be the major active ones related to the
hematinic effect of SPN. By comparing the chromatogram of SPN sample to
that of the mixed standard solution (Figure [122]7), peaks 10 and 12
were identified to be ginsenosides Rk[3] and 20(S)-Rg[3], respectively,
which had the major role in the hematinic effect of SPN.
FIGURE 7.
FIGURE 7
[123]Open in a new tab
The (A) chromatograms of SPN sample (black) and the mix standard
solution (red), and structures of (B) ginsenosides Rk[3] and (C)
20(S)-Rg[3]. Peak 10 and 12 correspond to ginsenosides Rk[3] and
20(S)-Rg[3], respectively. SPN, steamed Panax notoginseng.
According to the results of network pharmacology analyses, the two
constituents were also predicted to be the active ones. Therefore,
ginsenosides Rk[3] and 20(S)-Rg[3] were determined to be the target
constituents for the further investigation of their hematinic effect.
Experimental Validation for the Predicted Results
Validation for the Screened Active Constituents
After intraperitoneal injected ginsenosides Rk[3] and 20(S)-Rg[3] for
15 days, the quantities of WBC, RBC, Hb, and PLT from the peripheral
blood of mice were shown in Figures [124]8, [125]9, respectively.
Compared with the control group, the levels of WBC, RBC, Hb, and PLT in
the model group were significantly decreased (P < 0.01), indicating the
anemia model was successfully established. Compared with the model
group, the levels of WBC, RBC, Hb, and PLT after treated three doses of
ginsenoside Rk[3] were all increased. Meanwhile, the levels of WBC,
RBC, and PLT in the high-dose group were significantly increased (P <
0.05), suggesting that Rk[3] could reverse the decrease of the
quantities of WBC, RBC, Hb, and PLT in a dose-dependent way. For the
treatment with ginsenoside 20(S)-Rg[3] of three doses, the levels of
WBC, RBC, Hb, and PLT were all increased compared with the model group.
Besides, the levels of WBC and Hb in the moderate-dose and high-dose
groups were significantly increased (P < 0.05); and the levels of RBC
and PLT in the high-dose group were significantly increased (P < 0.05),
suggesting that the ginsenoside 20(S)-Rg[3] could reverse the decrease
of the quantities of WBC, RBC, Hb, and PLT in a dose-dependent way.
FIGURE 8.
FIGURE 8
[126]Open in a new tab
The contents of (A) WBC, (B) RBC, (C) Hb, and (D) PLT in the blood of
mice after treated with ginsenoside Rk[3]. Each value represents means
± SD (n = 10). ^∗P < 0.05 and ^∗∗P < 0.01, compared with the control
group; ^△ P < 0.05 and ^△ △ P < 0.01, compared with the model group.
FEJ, Fufang E’jiao Jiang; Hb, hemoglobin; PLT, platelet; RBC, red blood
cell; SD, standard deviation; WBC, white blood cell.
FIGURE 9.
FIGURE 9
[127]Open in a new tab
The contents of (A) WBC, (B) RBC, (C) Hb, and (D) PLT in the blood of
mice after treated with ginsenoside 20(S)-Rg[3]. Each value represents
means ± SD (n = 10). ^∗P < 0.05 and ^∗∗P < 0.01, compared with the
control group; ^△ P < 0.05 and ^△ △ P < 0.01, compared with the model
group. FEJ, Fufang E’jiao Jiang; Hb, hemoglobin; PLT, platelet; RBC,
red blood cell; SD, standard deviation; WBC, white blood cell.
Validation for the Predicted Target Proteins
As shown in Figure [128]10, the level of FECH in the model group was
significantly decreased compared with the control group (P < 0.05),
suggesting that the model was well established. Compared with the model
group, the levels of FECH and heme were all significantly increased
after the administration of different doses of Rk[3] and 20(S)-Rg[3] (P
< 0.05). In general, the levels of FECH and heme in livers of mice
treated with the middle-dose Rk[3] and 20(S)-Rg[3] were relatively
higher than those treated with low and high doses of drugs. The results
indicated that ginsenosides Rk[3] and 20(S)-Rg[3] had a positive effect
on improving the levels of FECH and heme, which was consistent with the
predicted results of network pharmacology analyses.
FIGURE 10.
FIGURE 10
[129]Open in a new tab
The contents of (A) FECH and (B) heme in the livers of mice. Each value
represents means ± SD (n = 10). ^△ P < 0.05, compared with the control
group; ^∗P < 0.05 and ^∗∗P < 0.01, compared with the model group. FECH,
ferrochelatase; FEJ, Fufang E’jiao Jiang; SD, standard deviation.
Discussion
Herbal medicines exert their therapeutic effects through the
synergistic effects of multiple constituents and targets. PN in raw and
steamed forms are historically supposed to be different in the
efficacies. [130]Lau et al. (2009) reported that the bleeding time of
rats treated with raw PN was shorter than those treated with SPN.
[131]Zhou et al. (2014) found that SPN could significantly increase the
levels of Hb and WBC, as well as the organ index of mice with BDS
induced by CTX, which were unconspicuous when treated with raw PN.
Based on our previous studies ([132]Xiong et al., 2017b), there was a
significant variation in the chemical composition between the two forms
of PN, which leaded to the difference in the pharmacologic effects of
raw and steamed PN. As shown in Figure [133]3, the levels of blood
routine parameters of mice treated with SPN were significantly
increased compared with the model group, which were also obviously
higher than those of mice treated with raw PN. The result was
consistent with the traditional use of SPN as a tonic to enrich the
blood.
Currently, methods for uncovering active constituents of herbal
medicines treating diseases mainly rely on retrospective analyses.
However, this method depends on large consumption of manpower and
material resources, which hinders the development of drugs. To address
this issue, we have developed firstly a more comprehensive approach
that integrates anemia-SPN networks to effectively discover potential
active constituents and targets involved. Technically, the prediction
accuracy of the drug targets and the completeness of the databases are
important to the method and will affect the creditability of the final
results. Therefore, we tried to reduce the false positive cases, such
as threshold filtering with the fit score in the drug target
prediction, significance analyzing with hypergeometric distribution
approach in disease targets enrichment, and reasonable topologic
parameter screening in the analyses of network ([134]Wang et al.,
2018). In addition, the combined prediction of multivariate data
analyses and verification of pharmacologic tests confirmed the
credibility of the model. The predicted results indicated that 14
constituents of SPN were interacted with 11 targets related to anemia
in the network. As shown in Table [135]1, many candidate proteins were
targeted by more than one compound. It suggested that these targets
might play an important role in the hematopoiesis, the modulation of
which could lead to the stimulation of various cytokines in the
hematopoietic microenvironment, enhancement of the function of internal
free radical scavenging system, facilitation of the absorption and
utilization of iron, improvement of the bone marrow hematopoietic
microenvironment, etc. ([136]Wang et al., 2012b; [137]Liu M. et al.,
2014). The common cross-targets shared by multiple constituents implied
that SPN might exert the synergistic therapeutic effect on anemia,
which was probably more effective than single compound. This suggested
that the herbal medicine might act on polypharmacological level, rather
than on one specific protein in order to combat complex diseases, such
as anemia.
“Degree,” “betweenness centrality,” and “closeness centrality” are
three key topological parameters that characterize the most influential
nodes in a network. According to [138]Li et al. (2007), if the “degree”
of a node was more than twofolds of the median degree of all nodes in a
network, such gene or protein was believed to play a critical role in
the network structure, and it could be treated as a hub gene or a hub
protein. “Betweenness centrality” was one of the significant indicators
of network essentiality because proteins with high betweenness were
essential for the functioning of the system by serving as a bridge of
communication between several other proteins in the network ([139]Melak
and Gakkhar, 2015). And “closeness centrality” was another one of the
significant indicators of network essentiality which represented the
average length of the shortest paths to access all other proteins in
the network. The higher the value, the more central the protein
([140]Zhuang et al., 2015). Therefore, we used the above parameters to
determine the importance of active constituents and action targets (the
nodes in our network), as well as the extent of their influence on the
spread of information through the network ([141]Tang et al., 2015).
Among the 11 predicted targets involved in the pathogenic process of
anemia, FECH was shown relatively higher values of degree and closeness
centrality, and was reported to be closely related to the production of
heme. According to Figure [142]11, FECH is the terminal heme synthesis
enzyme to catalyze the insertion of the imported iron into
protoporphyrin IX to produce heme. Gene mutation in FECH may cause
pathological changes like erythropoietic protoporphyria, an autosomal
dominant disease which can develop into cholelithiasis and varying
degrees of liver diseases ([143]Casanova-González et al., 2010). It was
reported that FECH forms an oligomeric complex with Mfrn1 and Abcb10 to
synergistically integrate mitochondrial iron importation for heme
biosynthesis ([144]Chen et al., 2010). Since heme is an important raw
material for hemoglobin synthesis and the increased heme level can
resulted in a significant enhancement of human hemoglobin production
([145]Liu L. et al., 2014), the variation of FECH and heme could be
investigated to verify the hemopoiesis induced by active constituents
of SPN.
FIGURE 11.
FIGURE 11
[146]Open in a new tab
The metabolic pathway of FECH and heme. ALA, δ-aminolevulinic acid;
ALAD, δ-aminolevulinic acid dehydrase; ALAS, δ-aminolevulinic acid
synthase; CPGIII, coproporphyrinogen III; CPGIIIOD, coproporphyrinogen
III oxidized decarboxylase; FECH, ferrochelatase; Gly, glycine; Hb,
hemoglobin; PBG, porphobilinogen; PBGD, porphobilinogen deaminase;
PPIX, protoporphyrinIX PPGIX, protoporphyrinogenIX; PPGIXO,
protoporphyrinogenIX oxidase; Suc-CoA, succinyl–coenzyme A; UPGIX,
uroporphyrinogen III; UPG?D, uroporphyrinogen III decarboxylase.
From Table [147]1, the majority of compounds were linked with more than
one target, indicating that these compounds might play the therapeutic
effect by acting on multi-targets. Among the 14 compounds in our
network, several of them might be essential. For example, in our
previous work ([148]Xiong et al., 2017a), we found the levels of some
saponins in PN were increased along with the steaming time and
temperature. Among them, ginsenosides Rh[4], Rk[3], 20(R)-Rg[3], and
20(R)-Rh[1] with higher contents or exclusively existed in SPN showed
higher contributions to the activities of SPN. The result was
consistent with the prediction in this research, that seven
constituents of ginsenosides Rh[4] (10), Rk[3] (11), 20(R)-Rg[3] (14),
20(S)-Rg[3] (15), 20(R)-Rh[1] (16), and 20(S)-Rh[1] (17), and
notoginsenoside R[2] (20) were predicted to be interacting with the
target of FECH for the treatment of anemia. It indicated that the
network pharmacology approach had great potential to identify active
constituents and alternative targets for the mechanism understanding
and drug development of herbal medicines.
To further determine the active constituents, the analysis of
fingerprint-effect relationship has been applied to screen
characteristic constituents related to the hematinic effect of SPN.
Multivariate data analyses such as PLSR and CCA are often used to
specify a linear relationship between a set of dependent variables from
a large set of independent variables, especially when the sample size
is small relative to the dimension of these variables
([149]Garza-Juárez et al., 2011; [150]Wu et al., 2015). According to
the results, Rk[3] and 20(S)-Rg[3] were predicted to be the major
bioactive constituents of SPN treating anemia, which were also included
in the prediction of network pharmacology analyses. The two
constituents were reported to own various protective effects in
previous studies. For example, 20(S)-Rg[3] could prevent the
progression of renal damage ([151]Kang et al., 2010), protect against
benzo[a]pyrene-induced genotoxicity in human cells ([152]Poon et al.,
2012), and protect against lipopolysaccharide-induced oxidative tissue
injury in the liver of rats ([153]Kang et al., 2009). And ginsenoside
Rk[3] was shown a protective effect against hypoxia-reoxygenation
induced H9c2 cardiomyocytes damage, which was often used as a major
ingredient of the compound preparation for ischemic heart diseases
([154]Sun et al., 2013). To validate the predicted results, the effects
of Rk[3] and 20(S)-Rg[3] on levels of blood routine parameters were
investigated based on the BDS model. Compared to the model group, the
high-dose Rk[3] and 20(S)-Rg[3] could significantly increase the levels
of WBC, RBC, and PLT in a dose-dependent way. The high-dose 20(S)-Rg[3]
also made a significant difference on improving the content of Hb. It
indicated that the two constituents had positive effect on improving
the BDS of mice, which was consistent with the predicted results of
network pharmacology and multivariate data analyses. Besides, the
levels of FECH and heme could be increased by the treatment of Rk[3]
and 20(S)-Rg[3] (Figure [155]10), suggesting that the two constituents
exert the hematinic effect by regulating the predicted target and its
downstream protein. Meanwhile, we noticed that the high-dose of
ginsenosides Rk[3] and 20(S)-Rg[3] also inhibited the production of
FECH and heme. That might be due to the bidirectional adjustment
between FECH and heme that FECH was an integral factor for the
biosynthesis of heme, whereas an excessive level of heme could inhibit
the production of FECH ([156]Dailey and Fleming, 1983; [157]Wu et al.,
2016). These results also indicated that the metabolism of FECH and
heme was involved in the development of anemia, of which the
stabilization could be regulated by SPN to resist hemolysis. However,
further studies are needed to understand the precise nature of these
contributing factors.
Conclusion
To unveil the bioactive constituents and investigate the action
mechanism of SPN for improving BDS, the network pharmacology approach
coupled with multivariate data analyses were performed. In this study,
we firstly predicted the active constituents and potential targets of
SPN related to the treatment of anemia disease. The results showed that
ginsenosides Rk[3] and 20(S)-Rg[3] were active constituents related to
the hematinic effect of SPN, which acted on the targets of FECH and
heme to improve the BDS. Although there could be various pathogens
causing the incident of anemia and only the hemolytic type was
investigated in the research, it also indicated potential areas for
further research of SPN as a botanical remedy for the treatment of
related diseases. The strategy employed does not only provide new
insights for a deeper understanding of the chemical basis and
pharmacology of SPN, but also demonstrate an efficient method for
potential discovery of drugs originating from herbal medicines.
Additional study on the therapeutic effect of SPN on other types of
anemia and the involved pathways will be further carried out.
Author Contributions
YX wrote this paper and carried out parts of data analyses. YH
constructed the network and verified the predicted targets. LC did the
multivariate data analyses and parts of pharmacologic tests. ZZ and YZ
conducted parts of the pharmacologic tests. MN provided the technical
support of network pharmacology. YX and XC supervised the project. All
authors read and approved the final manuscript.
Conflict of Interest Statement
The authors declare that the research was conducted in the absence of
any commercial or financial relationships that could be construed as a
potential conflict of interest.
Abbreviations
APH
acetylphenylhydrazine
BDS
blood deficiency syndrome
CCA
canonical correlation analysis
CTX
cyclophosphamide
FECH
ferrochelatase
FEJ
Fufang E’jiao Jiang
Hb
hemoglobin
OMIM
Online Mendelian Inheritance in Man
PLSR
partial least squares regression
PLT
platelet
PN
Panax notoginseng
PPI
protein–protein interaction
RBC
red blood cells
SPN
steamed Panax notoginseng
TCM
traditional Chinese medicine
TDT
TCM Database@Taiwan
WBC
white blood cells
Funding. This work was supported by National Natural Science Foundation
of China (81660661), Kunming University of Science and Technology
(KKSY201526065), and Yunnan Applied Basic Research Project (2016FD040).
^1
[158]http://lilab.ecust.edu.cn/pharmmapper/index.php
^2
[159]https://david.ncifcrf.gov/
Supplementary Material
The Supplementary Material for this article can be found online at:
[160]https://www.frontiersin.org/articles/10.3389/fphar.2018.01514/full
#supplementary-material
[161]Click here for additional data file.^ (44.2KB, docx)
References