Understanding ZHENG in traditional Chinese

medicine in the context of neuro-endocrine-immune

network
S. Li, Z.Q. Zhang, L.J. Wu, X.G. Zhang, Y.D. Li and Y.Y. Wang
Abstract: Traditional Chinese medicine uses ZHENG as the key pathological principle to understand the human homeostasis and guide the applications of Chinese herbs. Here, a systems biology
approach with the combination of computational analysis and animal experiment is used to investigate this complex issue, ZHENG, in the context of the neuro-endocrine-immune (NEI) system. By
using the methods of literature mining, network analysis and topological comparison, it is found
that hormones are predominant in the Cold ZHENG network, immune factors are predominant
in the Hot ZHENG network, and these two networks are connected by neuro-transmitters.
In addition, genes related to Hot ZHENG-related diseases are mainly present in the cytokine
cytokine receptor interaction pathway, whereas genes related to both the Cold-related and
Hot-related diseases are linked to the neuroactive ligand-receptor interaction pathway. These computational ndings were subsequently veried by experiments on a rat model of collagen-induced
arthritis, which indicate that the Cold ZHENG-oriented herbs tend to affect the hub nodes in the
Cold ZHENG network, and the Hot ZHENG-oriented herbs tend to affect the hub nodes in the
Hot ZHENG network. These investigations demonstrate that the thousand-year-old concept of
ZHENG may have a molecular basis with NEI as background.

Introduction

Traditional Chinese medicine (TCM) is a system with its

own rich tradition and over 3000 years of continuous
practice and renement through observation, testing and
critical thinking. TCM can be characterised as holistic
with emphasis on regulating the integrity of the human
body and the interaction between human individuals and
their environments. TCM applies multiple natural
therapeutic methods for patient management, with herbal
formulas as typical treatments.
ZHENG is the basic unit and the key concept in TCM
theory. All diagnostic and therapeutic methods in TCM
are based on the differentiation of ZHENG, and this
concept has been used for thousands of years in China
[1]. ZHENG can be seen as the TCM theoretical abstraction
of the symptom proles of a disease, not simply an assemblage of disease symptoms. Also, ZHENG is used as a
guideline in disease classication in TCM. For example,
patients suffering from the same disease may be categorised
into different ZHENGs, whereas different diseases may be
categorised as the same ZHENG. The Cold ZHENG
(HAN ZHENG in Mandarin) and Hot ZHENG (RE
ZHENG in Mandarin) are the two key statuses of
ZHENG, which therapeutically direct the use of Chinese
# The Institution of Engineering and Technology 2007
doi:10.1049/iet-syb:20060032
Paper rst received 12th March and in revised form 26th July 2006
S. Li, Z.Q. Zhang, L.J. Wu, X.G. Zhang and Y.D. Li are with the Bioinformatics
Division, TNLIST and Department of Automation, Tsinghua University,
Beijing 100084, Peoples Republic of China
Y.Y. Wang is with the China Academy of Chinese Medicine Sciences, Beijing
100700, Peoples Republic of China
E-mail: [email protected]
IET Syst. Biol., 2007, 1, (1), pp. 51 60

herbs in TCM. The terms of major symptom proles for

Cold and Hot ZHENGs are listed in Table 1 and a detailed
description is given in Supplementary 1. Correspondingly,
many Chinese herbs are categorised as either Hot-Cooling
type or Cold-Warming type: Hot-Cooling herbs are used
to remedy Hot ZHENG, and Cold-Warming herbs are
used to remedy Cold ZHENG.
Owing to the importance of ZHENG in both the theory
and the practice of TCM, much effort has been paid to
investigate its biomedical fundaments. Ou et al. [2] compared the antioxidation oxidation balance in biochemical
processes with the YIN YANG balance, an ancient
Chinese philosophy that underlies the concept of ZHENG
in TCM [1]. Ko et al. [3] reported that the pharmacological
basis of YANG-invigoration in Chinese medicine might
be primarily due to the enhancement of mitochondrial
ATP generation. In spite of all these efforts, it is still far
from a full understanding of the inherent mechanism of
ZHENG. And as for the methodology, it is known that the
scientic evaluation of TCM is very difcult using existing
conventional methods [4, 5].
The objective of this work is to explore the molecular
basis of ZHENG within the context of neuroendocrine-immune (NEI) system. In modern Western medicine (WM), NEI system acts as a pivot in modulating host
homeostasis and naturally optimising health through
complex communications among chemical messengers
(CMs), including hormones, cytokines and neurotransmitters [6, 7]. If we consider CMs as the biochemical
ingredients of the NEI system, then those genes that
(directly or indirectly) encode these CMs can be considered
as the genic ingredients of the NEI system.
Cold ZHENG and Hot ZHENG are widely applied in the
diagnosis and the treatment of patients suffering from
inammation, infection, stress and autoimmune disorders
51

Table 1: NEI-related terms and major symptom prole terms of Cold ZHENG and Hot ZHENG
Subjects

Terms (keywords)

Cold-ZHENG-related symptom

Cold (chill, coldness); cold pain; tastelessness; clear abundant urine (clear urine in large amounts);

prole terms
Hot-ZHENG-related symptom
prole terms

loose stool; pale tongue; white fur (white moss); tight pulse (stringy pulse)
Fever; heat (hot); diaphoresis; ushed face; burning pain; deep-coloured urine; red eyes; thirst;
desire for drinking; constipation; red tongue; dry tongue; thin fur (thin moss); yellow fur (yellow
moss); rapid pulse

NEI-related terms

Neuro-endocrine-immune; nerve-endocrine-immune; hypothalamic-pituitary; hypothalamus

pituitary; neuro-endocrinology; neuro-immunology; neuro-immunomodulation;
immune-neuroendocrine; endocrine-immune; neuroimmunoendocrinology;
psychoneuroimmunology

[8, 9]. According to WM, most of such disorders are related

to the bi-directional communication between the neuroendocrine and the immune systems [10]. And defects in
NEI will result in various kinds of pathologies [11, 12].
Thus, two aspects make NEI an ideal context to explore
the underlying molecular basis of ZHENG and to bridge
the gap between the ancient TCM and the modern medicine.
First, the NEI system acts as the host homeostasis mediator
during the course of various body disorders. Second, there
are some clues that patients with Cold ZHENG and Hot
ZHENG, two representative and mutually controlled
ZHENGs, present abnormal NEI functions [8, 9, 13]. For
example, patients suffering from rheumatoid arthritis (RA)
are found to experience an alteration of the NEI system
[14, 15]. And in TCM, most RA patients can be categorised
as either Cold-ZHENG-related RA or Hot-ZHENG-related
RA, and treated by the Cold-Warming or Hot-Cooling
TCM herbal formulas, respectively [9].
The vast amount of medical and biological literatures are
a huge reservoir of human knowledge on complex life
systems, but most of them exist in the form of pieces of isolated knowledge, for example the function of a single gene
or the interaction of a small number of molecules. Literature
mining techniques can help to build or to recover the whole
picture buried in the isolated reports, for example building
biological entity-relation networks from the literatures and
then generating hypothesis [16, 17]. In this paper, we
study the networks of both CMs and genes of NEI
systems with regard to Cold ZHENG and Hot ZHENG,
respectively, by using a co-occurrence literature mining
approach [16 18]. We compare the networks built for
Cold ZHENG and Hot ZHENG, and nd interesting features of these two ZHENGs in the NEI system, showing a
putative correspondence between ZHENGs and different
aspects of the NEI system. Some key factors in the Cold
and the Hot NEI networks are identied. Next, we selected
two groups of diseases with their typical symptom proles
corresponding to Cold ZHENG and Hot ZHENG, respectively, in the view of TCM. Gene investigation about
ZHENG-related diseases is performed to validate patterns
derived from the literature mining approach. Finally, as
RA has a close relationship with the NEI system and is
widely studied in both WM and TCM [9, 14, 15], we
choose RA as a disease model to explore ZHENG within
the context of the NEI system. Unfortunately, the direct biological measurement of ZHENG is hardly available.
Instead, we conducted experiments on the rat model of
collagen-induced arthritis (CIA, the most widely used
model for RA) [19, 20] to study the effects of the
Cold-Warming and Hot-Cooling TCM herbal formulas
(CWHF and HCHF, respectively) on the key CMs of the
NEI network.
52

2
2.1

Computational and experimental methods

Building the NEI networks

We summarised the synonyms of NEI by querying Medline

MeSH (medical subject headings) and excluding the irrelevant subheadings and postxal terms. Eleven available
keywords (Table 1) are used to search the PubMed database
(http://www.ncbi.nlm.nih.gov/entrez/query.fcgi). A total of
21 222 PubMed abstracts were downloaded, called the
NEI PubMed pool. The 20 327 non-redundant gene
symbols from the Human Genome Organisation (HUGO)
glossary (http://www.gene.ucl.ac.uk/nomenclature/) are
used to search for NEI-related genes and their relations.
Moreover, a dictionary of NEI-related CMs, that is hormones, cytokines and neuro-transmitters, are manually collected from a total of 261 full-text review articles in English
during the past 5 years (from September 30 2000 to
September 30 2005) by using hormone, cytokine and
neuro-transmitter as keywords, respectively, and
disease as the constrained condition to query the above
NEI PubMed pool. This dictionary contains 109 CMs (corresponding to 242 terms, abbreviations and synonyms, see
Supplementary 2). Using the HUGO genes and the
NEI-related CMs, we build a gene network and a CM
network for the NEI system via the co-occurrence literature
mining approach, which assumed that when two biological
entities are co-cited in the same text unit there should be a
potential biological relationship between them [16 18].
The Neato program in the Graphviz software (AT&T;
http://www.research.att.com/sw/tools/graphviz/) is adopted
to visualise these networks.

2.2

Building the ZHENG networks

Keywords for Cold and Hot ZHENGs are collected

according to the ZHENG symptom proles dened by
the authoritative and standard TCM terminology
(Table 1) [21, 22]. We eliminated the homonymous
terms that contain the word hot or heat but are unrelated to Hot ZHENG. For example, Hot Shock Protein
and Heat Shock Protein are excluded. Using the Cold
ZHENG and the Hot ZHENG keywords, 189 and 382
NEI-related PubMed abstracts are collected, respectively.
Using the same method described in the NEI network
construction, the gene/CM-based Cold ZHENG network
and the Hot ZHENG network are built from their corresponding abstract pools. The function annotation of
these genes and CMs are extracted from the NCBI
database (http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?
CMD=search&DB=gene).
IET Syst. Biol., Vol. 1, No. 1, January 2007

2.3

Topological analysis of the ZHENG networks

The topological characteristic of node i in a network

is represented by a topology vector vi(W ) [ci1 ,
ci2 , . . . , cin]T, where n is the total number of unique nodes
in both Cold and Hot ZHENG networks, W fC, Hg
refers to the Cold or Hot ZHENG network and cij is the indicator variable of the association between nodes i and j
(cij 1 if i = j and there is an association; 0 otherwise).
The topology distance between Cold ZHENG and Hot
ZHENG networks is dened as di (vi(H) 2 vi(C))T. e,
where vi is the topology vector of node i, and e is the unit
vector with all elements equal to 1. We normalise the topology distance to the range of [21, 1] as
d i

viH  viC T  e
di

viH viC T  e viH viC T  e

As we classify a node i as a Hot ZHENG node if di . 0, or

as a Cold ZHENG node if di , 0, we thus call di the topological temperature for the convenience of discussion.
Then, we assess the observed Cold/Hot ZHENG proportion
within each category of NEI using the cumulative binomial
distribution.
 N
c0 
X
1
N
Pcc0
c
2
c0
where N is the total number of nodes in one category,
excluding those with di 0; c0 is the smaller number
between the node number of the Cold ZHENG nodes and
the node number of the Hot ZHENG nodes.
2.4 Construction of ZHENG-related disease data
sets and ZHENG-related NEI pathways
From Clinical Medicine Database (http://www.cintcm.com)
hosted by China Academy of Chinese Medicine Sciences,
where information about more than 4000 diseases was
described, we selected ZHENG-related diseases whose
properties can be typically differentiated and categorised
as either Cold ZHENG or Hot ZHENG. Genes associated
with the obtained 21 Cold ZHENG-related diseases and

38 Hot ZHENG-related diseases (Table 2) were collected

from the OMIM database (http://www.ncbi.nlm.nih.gov/
entrez/query.fcgi?db=OMIM). The NEI-related pathways
of these genes in KEGG (http://www.genome.jp/kegg/)
were obtained via the DAVID database (http://apps1.niaid.
nih.gov/david/). In the DAVID annotation system, Fishers
exact test was adopted to measure the gene-enrichment
in annotation terms (http://david.abcc.ncifcrf.gov/helps/
functional_annotation.html#sher). Here, Fishers exact
test is used to measure whether the proportion of genes in
a specic pathway for the query genes is signicantly
higher than that for the human genomic background genes.
2.5 Effects of ZHENG-oriented herbal treatments
on hub nodes
Experiments are conducted on CIA rats to investigate the
effects of the ZHENG-oriented herbal treatments, CWHF
and HCHF, on the hub nodes in the Cold ZHENG and the
Hot ZHENG networks. CWHF and HCHF (Table 3) have
been clinically proved to be effective to RA [23], and
these two formulas can also improve arthritis signs and
joint damages in CIA rats [24, 25]. In our experiments,
rats are randomly divided into four groups, namely
normal, CIA, HCHF-treated and CWHF-treated. The variations of the concentration levels in the four groups,
measured using radioimmunoassay at 6-h intervals from
0:00 (midnight) to 24:00, are recorded for a number of
important hub CMs in the Hot and Cold ZHENG networks.
Differences between the CIA rats and the normal rats, the
CIA rats and each of the two treatment rats were analysed
using the one-way ANOVA followed by Bonferroni post
hoc test. P , 0.05 was considered signicant. The experiments described were performed in accordance with the
UK Animals (Scientic Procedures) Act of 1986. Details
of the experiments are described in Supplementary 3.
3
3.1

Results and discussion

Computational NEI network

In the gene-based NEI network and the CM-based NEI

network constructed in Section 2.1, the genes/CMs are
the nodes in the networks, and an edge between two

Table 2: List of diseases related to Cold ZHENG and Hot ZHENG

Disease data sets

Number

Diseases

Cold ZHENG-related

21

Protein deciency; hypothyroidism; child pituitary dwarsm; amenorrhoea galactorrhoea

diseases

syndrome; over weakness; deciency and cold in large intestine; anaemic cardiopathy; adult
anterior pituitary hypofunction; hypopituitarism; chronic pulmonary heart disease; occlusive
arteriosclerosis; chronic renal failure; litchi disease; cretinism; neurosis; hyperprolactinaemia;
empty sella syndrome; hypothalamus syndrome; psychonosema associated with
hypothyroidism; impotence

Hot ZHENG-related

38

diseases

Infectious mononucleosis; rheumatic fever; epidemic haemorrhagic fever; typhoid fever;

paratyphoid; acute interstitial nephritis; acute pyelonephritis; acute enteritis; hyperthyroidism;
acute appendicitis; acute pancreatitis; acute cholecystitis; suppurative infection; acute thyroiditis;
acute peritonitis; trench fever; Australian tick-borne spotted fever; Q fever; Rocky mountain
spotted fever; miliary fever; Kyasanur forest fever; Omsk haemorrhagic fever; Lassa fever; yellow
fever; ratbite fever; child high-fever convulsion; kala-azar; epidemic haemorrhagic fever;
relapsing fever; relapsing febrile non-suppurative panniculitis; septicemia; acute hepatitis gravis;
epidemic cerebrospinal meningitis; epidemic encephalitis B; acute leukaemia; child acute
pancreatitis; child viral myocarditis; acute febrile dermatosis of neutrophilic granulocytosis

IET Syst. Biol., Vol. 1, No. 1, January 2007

53

Table 3: Composition of HCHF and CWHF

Formula

Herbs

Aims

HCHF

Ku-Shen (Radix sophorae

Hot ZHENG

averscentis)
Huang-Bai (Cortex phellodendri)
Qing-Feng-Teng (Caulis sinomenii)
Bi-Xie (Rhizoma dioscoreae
hypoglaucae)
CWHF

Fu-Zi (Radix aconiti lateralis preparata)

Cold ZHENG

Bai-Zhu (Rhizoma atractylodis

macrocephalae)
Gui-Zhi (Ramulus cinnamomi)
Juan-Bai (Herba selaginellae)

nodes indicates that these two genes or CMs are related. The
gene-based NEI network reects a view of the whole structure of NEI at genic level, including 1585 genes and 8161
relation edges. The CM-based NEI network reveals the
major functional component of NEI, which is composed
of 108 CMs and 1607 relation edges. The number of
edges a node has in a network is called the degree of that
node [26 28], which indicates how many genes/CMs one
gene/CM is related with. If the degree of a node is more
than 2 fold of the median degree of all nodes in a
network, such gene or CM is believed to play a critical
role in the network structure, and we treat it as a hub gene
or a hub CM [26 28]. As shown in Fig. 1a, the NEI
network, has the property of a scale-free network, in
which the connection degrees of the nodes follow a power
law. Recent study [29] shows that the scale-free networks
tend to contain centrally located, highly connected hub
nodes that have dramatic inuence on the way a network
operates. Therefore it is believed that the hub genes in the
network may play a key role in the NEI system. The
CM-based NEI network is built on the specic CM list
and therefore the power-law property is not observed.
Most of the hub nodes in the gene-based NEI network
also appear in the CM-based NEI network. As shown in
Table 4, these hub nodes include the tumour necrosis
factor (TNF), proopiomelanocortin [POMC, other designations as adrenocorticotropic hormone (ACTH) according
to NCBI database], interleukin (IL)-6, prolactin (PRL),
corticotropin releasing hormone (CRH) and so on. We
also found that the hub nodes in the networks mainly
belong to cytokines of the immune system as well as
hypothalamic pituitary (HP)-target organ axes such as
HP adrenal (HPA) and HP thyroid (HPT) related hormones of the neuro-endocrine system. According to the
schematic outline of the neuro-endocrine factors that regulate the secretion of the adrenal cortex [30], the hub nodes
and their connections identied in our study are representatives of the NEI system. ACTH (79dCM: degree 79 in our
CM-based network; 274dGene: degree 274 in our genebased network) is released from the pituitary and is stimulated synergistically by CRH (75dCM and 142dGene) and
arginine vasopressin (AVP, 58dCM and 102dGene). The
secretion of ACTH is inhibited at the pituitary level by circulating glucocorticoids (cortisol in human, 41dCM, and
corticosterone in rodents, 20dCM), which also regulate
CRH and AVP through negative feedback control. CRH
and AVP neurons are in turn subject to a wide range of
inuences circulating cytokines and many neurotransmitters. For example, CRH and AVP are stimulated
by acetylcholine (ACH, 32dCM and 8dGene), and are
54

Fig. 1 Computational neuro-endocrine-immune and ZHENG

systems
a Co-citation proles in the gene-based NEI network
b Gene-based Cold ZHENG network
c Gene-based Hot ZHENG network
Shown on a loglog plot, the connectivity of the gene-based networks
(NEI, Cold ZHENG and Hot ZHENG) follows a power law,
P(k) k 2g
CM-based networks identify approximately the same set of hub nodes
as in their corresponding gene-based networks

inhibited by gamma-amino butyric acid (GABA, 41dCM).

Moreover, each of the major inammatory cytokines, interleukin (IL)-1 (73dCM and 25dGene), IL-2 (53dCM and
101dGene), IL-6 (74dCM and 201dGene) and TNF-alpha
(146dCM and 315dGene), is capable of activating ACTH
release [30]. The peripheral immune system provides a
chemosensory system by which the presence of foreign
molecules can be communicated to the brain and thus can
induce an appropriate response [30].
IET Syst. Biol., Vol. 1, No. 1, January 2007

Table 4: Hub nodes of the gene-based and the CM-based networks for NEI, Cold ZHENG and Hot ZHENG
NEI network

Cold ZHENG network

Hot ZHENG network

(Number of co-citations)

(Number of co-citations)

(Number of co-citations)

Gene-based

CM-based

Gene-based

CM-based

Gene-based

CM-based

(Median

(Median

(Median

(Median

(Median

(Median

connectivity: 4)

connectivity: 24)

connectivity: 2)

connectivity: 3)

connectivity: 2)

connectivity: 3)

TNF (315)

TNF (146)

POMC (20)

ACTH (18)

TNF (19)

TNF-alpha (13)

POMC (274)

ACTH (79)

TRH (12)

TRH (14)

IL6 (10)

IL-6 (9)

IL6 (201)

IL-6 (74)

CORT (8)

AVP (8)

IL1RAP (4)

IL-1 (7)

PRL (156)

IL-1 (73)

CRH (7)

CORT (6)

CRH (142)

CRH (75)

AVP (5)

CRH (6)

TRH (112)

PRL (69)

AVP (102)

NF-kappaB (68)

IL2 (101)

AVP (58)

VIP (63)

IFN-gamma (57)

NPY (70)

VIP (56)

NA (69)

NA (54)

NOS2A (44)

TRH (54)

HTR1A (33)

IL-2 (53)

IL1B (25)

NO (50)

IGFBP7 (21)

IGF (49)

IFNA (20)

NPY (49)

NFKB (15)

5-HT (48)

3.2 Computational Cold ZHENG and Hot ZHENG

networks
Using the same method described earlier, we build a Cold
ZHENG network and a Hot ZHENG network from the
NEI-related PubMed abstracts with the standard Cold
ZHENG and Hot ZHENG keywords (Table 1), respectively.
For Cold ZHENG, the gene network is composed of
142 nodes with 120 edges, and the CM network is composed of 36 nodes with 69 edges. For Hot ZHENG, the
gene network has 202 nodes and 169 edges, and the CM
network has 43 nodes and 55 edges. Similar to the NEI
network, the degree distribution in the gene-based
ZHENG networks follows a power law (Figs. 1b and c),
and most hub nodes in the gene-based ZHENG network
are identied in the CM-based ZHENG network
(Table 4). The hub genes/CMs in the Hot ZHENG
network (TNF/TNF-alpha, -beta, IL6/IL-6, and IL1RAP/
IL-1) belong to the category of immune cytokines,
whereas the hub genes/CMs in the Cold ZHENG network
are characterised as hypothalamus pituitary hormones
(POMC/ACTH, CRH/CRH, TRH/TRH and CORT/
CORT) or neuro-transmitters (AVP/AVP).
Fig. 2a shows the combined CM network of Cold and Hot
ZHENGs, where two networks are fused through the nodes
and edges shared by both ZHENGs. The nodes appearing in
Hot ZHENG network are shown in red, those in Cold
ZHENG network are shown in blue and those appearing
in both networks are shown in purple. A Yin-Yang map is
drawn to disern the main bodies of two networks and indicate a mutual transformation of both may occur in given
changes. It can be seen that in the Cold ZHENG network,
HPA hormones such as cortisol, corticosterone, CRH and
ACTH are closely interconnected; HPT hormones such as
thyroid hormone (T) 3, T4, throtropin-releasing hormone
(TRH) and thyroid-stimulating hormone (TSH) are also
closely interconnected. In the Hot ZHENG network,
IET Syst. Biol., Vol. 1, No. 1, January 2007

cytokines such as TNF-alpha, TNF-beta, IL-6 and IL-1

are closely interconnected.
Fig. 2b displays the topological temperature of the CMs
in the Hot and the Cold ZHENG networks. On the basis of
cumulative binomial distribution, the CMs such as ACTH,
TRH, TSH, GH and GIH secreted by hypothalamus or pituitary are mainly involved in the Cold ZHENG network
(P 0.0059), immune factors such as cytokines of
TNF-alpha, IL-6, and IL-1 are mainly involved in the Hot
ZHENG network (P 2.1  1026) and neuro-transmitters
such as catecholamine are distributed in both ZHENG networks (P 0.254). Fig. 2c shows the topological temperature of the genes presented in both ZHENG networks. It is
observed that the genes in the gene-based Cold ZHENG and
Hot ZHENG networks exhibit similar patterns as in the
CM-based networks. Hormone-related genes are predominant in the Cold ZHENG network (P 0.046),
immune-related genes are predominant in the Hot
ZHENG network (P 0.00047) and neuro-transmitterrelated genes are distributed in both networks (P 0.5).
These observations provide evidence that the ZHENG
concept has its molecular base and different ZHENGs
reect their characteristics in different aspects of the
complex NEI system.
3.3 ZHENG-related diseases and ZHENG-related
NEI pathways
The typical symptoms of Cold ZHENG are chill without
fever, whereas the typical symptoms of Hot ZHENG are
fever without chill [21, 22]. We selected 59 diseases
representing the major symptom features ZHENGs and
divided them into two groups corresponding to either
Cold ZHENG or Hot ZHENG. The Hot group contains
38 diseases, whereas the Cold group contains 21 diseases (Table 2). From the OMIM database, a total of
201 genes related to Cold ZHENG-related diseases and
55

Fig. 2

Network patterns of ZHENG

a CM-based ZHENG network

Colour of a node signies its ZHENG state (blue, Cold ZHENG; red, Hot ZHENG; purple, both Cold and Hot ZHENGs. The same to b)
Network can be divided into two major network clusters: Cold ZHENG network (composed of all Cold ZHENG-related nodes), and Hot ZHENG
network (composed of all Hot ZHENG-related nodes)
The Yin-Yang background here is to indicate the possible relationship between Cold-Hot ZHENG networks and Yin-Yang theory of TCM, a mutual
transformation of both may occur in given quantitative changes
b Topological temperature (see methods) of each nodes in the CM-based Cold ZHENG (210) and Hot ZHENG (0 1) networks
c Gene-based Cold ZHENG (21 0) and Hot ZHENG (0 1) networks
Centre of the CM name or gene name corresponds to its topological temperature

603 genes related to Hot ZHENG-related diseases are collected, among which 60 genes are associated with both
sets. By investigating the NEI-related KEGG pathways
of these genes via the DAVID database, we found that
in contrast to genes specic to Cold diseases, genes
specic to Hot diseases are signicantly enriched in
the cytokine cytokine receptor interaction pathway
(P 5.9  1025 for Hot diseases and P 0.225 for
Cold diseases). The relationship between genes in
Cold diseases and hormones are not identied since
KEGG contains fewer pathways for complex hormone
regulations. These results are in agreement with the biological patterns we observed in the Cold and Hot
ZHENG networks (see previous section). Moreover, the
56

results summarised in Table 5 show that genes shared

by both Hot and Cold diseases are signicantly
enriched in the neuroactive ligand receptor interaction
pathway (P 2.21  1027). Together with the network
structure of ZHENG in Fig. 2, we will speculate that
the neuro-transmitters can be the common factors shared
by both Cold ZHENG and Hot ZHENG.
3.4 Effects of ZHENG-oriented herbal treatments
on the network hub nodes
Effects of the ZHENG-oriented herbal treatments on the hub
nodes of ZHENG networks are evaluated using rat CIA, an
experimental model of human RA [19, 20]. As shown in
IET Syst. Biol., Vol. 1, No. 1, January 2007

Table 5: NEI-related KEGG pathways for genes in Cold ZHENG and Hot ZHENG-related diseases
Genes

Number of genes

Genes related to Cold

151

ZHENG-related diseases only

Genes related to Hot

543

ZHENG-related diseases only

Genes related to both diseases

60

NEI-related KEGG pathway (Homo sapiens)

P-valuesa

Neuroactive ligand receptor interaction

0.180

Cytokinecytokine receptor interaction

0.225

Neuroactive ligand receptor interaction

0.030

Cytokinecytokine receptor interaction

5.9  1025

Neuroactive ligand receptor interaction

2.21  1027

P-values are calculated by Fisher exact test in DAVID. Fishers exact test is used to measure whether the proportion of genes in a
specic pathway for the query genes is signicantly higher than that for the human genomic background genes

Table 3, we adopted a CWHF (which is used to remedy Cold

ZHENG in TCM) to represent Cold ZHENG-oriented herbal
treatment, and a HCHF (which is used to remedy Hot
ZHENG in TCM) to represent Hot ZHENG-oriented herbal
treatment in our experiment. Circulating concentrations of
two hub nodes in the Hot ZHENG network, namely cytokines of TNF-alpha and IL-6, and two hub nodes in the
Cold ZHENG network, namely HPA hormones of ACTH
and corticosterone, are measured every 6 h during the 24-h
period of the experiment.
Cytokines and HPA hormones are two key players in the
classical neuro-endocrine feedback loop [11, 31, 32], where
both TNF-alpha, a cytokine at the top of the inammatory
cascade [33], and IL-6, a key pathogenic element in inammatory succeed, are potential activators of the HPA axis
[34 36]. During inammation, TNF-alpha is secreted rst
[37], IL-1beta and IL-6 successively [35, 36]. It is found
that the cytokine system and the HPA axis are disorderly
in many NEI-related diseases, including RA [32, 38]. In
our study, as shown in Fig. 3, the secretion of serum cytokines increased signicantly in rat CIA, for example
TNF-alpha is increased at 24:00 (P 0.043), IL-6 is
increased at 6:00 (P 0.048), 18:00 (P 0.021) and
24:00 (P 0.002) (ANOVA and Bonferroni test, similarly
hereinafter). The over-production of TNF-alpha and IL-6
indicates the progression of arthritis in CIA rats 33 days
after immunisation. Similar to the human RA patients,
where normal ACTH secretion and abnormal corticosterone
secretion are found [39], the secretion of corticosterone in
the CIA rats evidently increases at 6:00 (P 0.039) and
decreases at 18:00 (P 0.037), implying an adrenal disorder, which may be due to the stimulation of inammatory
cytokines such as TNF-alpha and IL-6 [32, 40]. Note that
the normal behaviour of ACTH in CIA rats may be a
result of the sparse sampling times. Some researchers
consider that cytokines such as IL-6 also act directly on
the adrenal cortex to stimulate corticosterone secretion
[11, 41].
The production of both cytokines and hormones is known
to exhibit diurnal rhythm in human RA and its rat models
[10, 32, 42]. Peak production of the proinammatory cytokines occurs at the time when plasma cortisol is at the
lowest [43]. In early untreated RA patients, it is found
that IL-6 precedes ACTH and cortisol by 1 and 2 h [40],
respectively. In our study, CIA rats present the similar
phenomena such that comparing with those of the normal
rats, the lower level of plasma corticosterone and the
higher level of serum IL-6 are observed at 18:00 simultaneously. The levels of both plasma corticosterone and
serum IL-6 increase signicantly at 6:00 clock in CIA
rats, suggesting that the HPA axis of CIA apparently is
insufcient to inhibit ongoing inammation, although
endogenous IL-6 may still stimulate the secretion of
IET Syst. Biol., Vol. 1, No. 1, January 2007

corticosterone. Thus, TNF-alpha and IL-6 stimuli (24:00)

are speculated to be responsible for the earlier peak of corticosterone (6:00) in CIA rats. The high levels of circulating
IL-6 in the CIA group from 18:00 to 6:00 form a continual
stimulation to the HPA axis, leading to adrenal insufciency, a similar condition as in humans [32]. Such results
demonstrate that CIA rats exhibit an impairment of the
HPA-cytokine feedback loop in which inammatory
cytokines are overproduced and the HPA axis is hyporesponsive [31].
In our ZHENG-oriented herbal treatment, it shows that
the HCHF-treatment can signicantly suppress the hub
nodes (e.g. TNF-alpha, IL-6) of the Hot ZHENG network
in rat CIA (Figs. 3a and b). For example, in the
HCHF-treatment group, signicant suppression can be
observed for IL-6 and TNF-alpha at 24:00 (IL-6,
P 0.003; TNF-alpha P 0.003 against CIA rats). In
addition, The HCHF treatment is observed to regulate the
abnormal secretion of corticosterone, for example the high
plasma corticosterone level of CIA at 6:00 (P 0.039,
CIA against normal rats) is reduced after the HCHF treatment (P 0.027, against CIA rats). As described earlier,
the inammatory cytokines TNF-alpha and IL-6 are the
potential activators of the HPA axis [34]. Thus, the
improvement of corticosterone secretion at 6:00 caused by
HCHF may be due to HCHFs ability to inhibit the production of TNF-alpha and IL-6 at 24:00, and then weaken
the stimulation of both cytokines to the HPA axis,
especially to the abnormal adrenal of CIA rats.
In contrast, the CWHF-treatment rats show suppression
of the hub nodes (e.g. ACTH, corticosterone) in the Cold
ZHENG network. For example, after CWHF treatment,
plasma corticosterone is dramatically inhibited at 6:00
(P 0.032) and 24:00 (P 0.046), and ACTH is inhibited
at 24:00 (P 0.002). TNF-alpha, a cytokine shared by both
ZHENG networks and highlighted in the Hot ZHENG
network (positive topological temperature), is also inhibited
by CWHF at 24:00 (P 0.002). It is known that synthetic
glucocorticoids, such as dexamethasone, can signicantly
suppress the rate of IL-6 gene transcription [44]. And the
component extracted from Fu-Zi (Radix aconiti lateralis
preparata), an herb in CWHF, is also found to affect the
HPA axis directly [45]. Therefore CWHF may have an
effect similar to exogenous hormones, leading to the
suppression of endogenous hormones ACTH and corticosterone in CIA rats.
From the above results, we speculate that HCHF tends to
affect the inammatory cytokines (hub nodes in Hot
ZHENG network) and then affect corticosterone through
the cytokine-HPA pathway (grey line arrow in Fig. 3b),
whereas CWHF tends to affect HPA axis hormones (hub
nodes in Cold ZHENG network) and then affect some of
cytokines through the HPA-cytokine feedback loop (black
57

Fig. 3

Effect of herbal formula on the different hub nodes of the ZHENG Network

a Effects of the CWHF and the HCHF on the four hub nodes of the ZHENG network within 24 h using CIA rats
b HCHF tends to affect the inammatory cytokines (hub nodes in Hot ZHENG network) then affect corticosterone through the cytokine-HPA
pathway (grey-line arrow)
CWHF tends to affect HPA axis hormones (hub nodes in Cold ZHENG network) then affect some of cytokines through the HPA-cytokine feedback
loop (black dot-line arrow)

dot-line arrow in Fig. 3b). Thus, with the combination of

computational analysis and experiment, our investigation
shows that the Cold ZHENG-oriented and Hot
ZHENG-oriented Chinese herbs modulate the HPAcytokine loop of CIA rats in different ways. CWHF tends
to affect the HPA hormones, the hub nodes in the computational Cold ZHENG network, whereas HCHF tends to
affect the immune cytokines, the hub nodes in the computational Hot ZHENG network. Such observations provide
an interesting biological explanation for the traditional
58

ZHENG management philosophy of Warming the Cold

and Cooling the Hot [1] in TCM.
4

Conclusions

TCM diagnosis and treatment are characterised by the

holistic view and the central concept of ZHENG. Here,
we performed a co-occurrence text mining to learn which
NEI-related genes/CMs participate in Cold ZHENG
network and Hot ZHENG network, respectively. Some
IET Syst. Biol., Vol. 1, No. 1, January 2007

relationships between Cold/Hot ZHENG and CMs/genes

in the NEI system are found and hypotheses are generated
by using this holistic method. Then, genes and corresponding NEI pathways from two groups of Cold ZHENG-related
and Hot ZHENG-related diseases are statistically evaluated.
The results validated the different biological patterns
identied in Cold ZHENG and Hot ZHENG networks.
We further experimentally conrmed that some key
players distinguishing Cold/Hot ZHENG are affected by
corresponding Cold/Hot ZHENG-oriented herbal treatments, respectively. As a result, this work highlights a possible molecular foundation for Cold ZHENG and Hot
ZHENG with regard to the NEI system: hormones may be
related to Cold ZHENG, immune factors may be related
to Hot ZHENG and they may be interconnected by neurotransmitters. Thus, Cold ZHENG and Hot ZHENG reect
two typical conditions of the internal imbalances of NEI,
and both of which should be taken into consideration
during disease diagnosis and treatment. Moreover, from a
methodological point of view, our work provides a new
approach towards understanding and rening the theory
and the principle of TCM within the framework of
modern science, which, in turn, can provide an alternate
perspective on the integrative NEI system. This preliminary
study is the rst step to understand ZHENG in viewing NEI
network. And we believe that uncovering the molecular
basis of ZHENG may help to design a tailored diagnosis
and treatment for patients in the future by evaluating their
ZHENG condition as well as the internal NEI imbalance,
and ZHENG-related approach is very likely to play a role
in the up-coming era of personalised medicine.
5

Acknowledgments

6
7

8
9
10
11

12
13
14
15
16
17

18

This work is supported by NSFC (30200365), the Key

Project of MOE (104009), the FANEDD (200366) and the
973 Program (2003CB517106 and 2004CB518605) of the
Republic
People's R. of
of China. We would like to express our great
appreciation to Dr. B. Li (Boston University, USA),
Dr. A. Lu (China Academy of TCM) and Mr. C. Zhang,
B. Zhang in our laboratory for helpful discussions and
comments.

19

23

Supplementary materials

Supplementary 1: explains the TCM terms about Cold

ZHENG and Hot ZHENG.
Supplementary 2: gives the manually collected dictionary of
NEI-related chemical messengers.
Supplementary 3: describes the experimental procedures for
CIA rats under the Cold/Hot ZHENG-oriented herbal
treatments.

20
21
22

24

25

26

References

1 Gu, C.D. (Ed.): The inner classic of the yellow emperor, essential
questions (Huangdi Neijing, Suwen) (Peoples Medical Publishing
House, Beijing, 1956), pp. 186191
2 Ou, B., Huang, D., Hampsch-Woodill, M., and Flanagan, J.A.: When
east meets west: the relationship between yin-yang and antioxidation
oxidation, FASEB J., 2003, 17, pp. 127129
3 Ko, K.M., Mak, D.H., Chiu, P.Y., and Poon, M.K.: Pharmacological
basis of Yang-invigoration in Chinese medicine, Trends
Pharmacol. Sci., 2004, 25, pp. 3 6
4 Xue, T., and Roy, R.: Studying traditional Chinese medicine,
Science, 2003, 300, pp. 7071
5 Wang, M., Lamers, R.J., Korthout, H.A., van Nesselrooij, J.H.,
Witkamp, R.F., van der Heijden, R., Voshol, P.J., Havekes, L.M.,
Verpoorte, R., and van der Greef, J.: Metabolomics in
the context of systems biology: bridging traditional Chinese
IET Syst. Biol., Vol. 1, No. 1, January 2007

27
28
29

30

31

medicine and molecular pharmacology, Phytother. Res., 2005, 19,

pp. 173 182
Besedovsky, H.O., and Sorkin, E.: Network of immuneneuroendocrine interactions, Clin. Exp. Immunol., 1977, 27, pp. 1 12
Roth, J., LeRoith, D., Shiloach, J., Rosenzweig, J.L., Lesniak, M.A.,
and Havrankova, J.: The evolutionary origins of hormones,
neuro-transmitters, and other extracellular chemical messengers:
implications for mammalian biology, N. Engl. J. Med., 1982, 306,
pp. 523 527
Hsu, C.H., Yu, M.C., Lee, C.H., Lee, T.C., and Yang, S.Y.: High
eosinophil cationic protein level in asthmatic patients with Heat,
Zheng, Am. J. Chin. Med., 2003, 31, pp. 277 283
Li, S.: Advances in TCM symptomatology of rheumatoid arthritis,
J. Tradit. Chin. Med., 2002, 22, pp. 137 142
Eskandari, F., Webster, J.I., and Sternberg, E.M.: Neural immune
pathways and their connection to inammatory diseases, Arthritis
Res. Ther., 2003, 5, pp. 251265
Haddad, J.J., Saade, N.E., and Saeh-Garabedian, B.: Cytokines and
neuro-immune-endocrine interactions: a role for the hypothalamic
pituitary adrenal revolving axis, J. Neuroimmunol., 2002, 133,
pp. 1 19
Dorner, G.: Hormone-dependent brain development and new
endocrine prophylaxis, Exp. Clin. Endocrinol., 1989, 94, pp. 4 22
Zhang, X., Ji, B., Chen, B., and Xie, Z.: Relationship of cytokines and
cold-heat syndrome differentiation in patients of duodenal ulcer,
Zhongguo Zhong Xi Yi Jie He Za Zhi, 1999, 19, pp. 267269
Wilder, R.L.: Neuroimmunoendocrinology of the rheumatic
diseases, past, present, and future, Ann. NY Acad. Sci., 2002, 966,
pp. 1319
Choy, E.H.S., and Panayi, G.S.: Cytokine pathways and joint
inammation in rheumatoid arthritis, N. Engl. J. Med., 2001, 344,
pp. 907 916
Jenssen, T.K., Lagreid, A., Komorowski, J., and Hovig, E.: A
literature network of human genes for high-throughput analysis of
gene expression, Nat. Genet., 2001, 28, pp. 2128
Zhang, C., and Li, S.: Modeling of neuro-endocrine-immune network
via subject oriented literature mining. Proc. 4th Int. Conf. on
Bioinformatics of Genome Regulation and Structure, Novosibirsk,
Russia, 2004, pp. 167170
Li, S., Wu, L.J., and Zhang, Z.Q.: Constructing biological networks
through combined literature mining and microarray analysis:
a LMMA approach, Bioinformatics, 2006, 22, pp. 2143-2150
Trentham, D.E., Townes, A.S., and Kang, A.H.: Autoimmunity to
type II collagen an experimental model of arthritis, J. Exp. Med.,
1977, 146, pp. 857 868
Staines, N.A., and Wooley, P.H.: Collagen arthritis what can it
teach us?, Br. J. Rheumatol., 1994, 33, pp. 798807
State Bureau of Technical Supervision: The clinic terminology of
traditional Chinese medical diagnosis and treatment ZHENG
(Standards Press of China, Beijing, 1997), p. 52
State Administration Bureau of TCM: The criteria of diagnosis and
therapeutic effect of diseases and ZHENGs in TCM (Nanjing
University Press, Nanjing, 1994), p. 29
Li, J.R., and Tong, X.L.: Discussion on the rheumatism (Bi Zheng
Tong Lun) (Anhui Science & Technology Publishing House, Hefei,
1987), p. 52
Li, S., Lu, A.P., and Jia, H.W.: Therapeutic actions of the Chinese
herbal formulae with cold and heat properties and their effects on
ultrastructures of synoviocytes in rats of the collagen-induced
arthritis, J. Tradit. Chin. Med., 2002, 22, pp. 290 296
Li, S., Lu, A.P., Wang, Y.Y., and Li, Y.D.: Suppressive effects of a
Chinese herbal medicine Qing-Luo-Yin extract on the angiogenesis
of collagen induced arthritis in rats, Am. J. Chin. Med., 2003, 31,
pp. 713 720
Song, C., Havlin, S., and Makse, H.A.: Self-similarity of complex
networks, Nature, 2005, 433, pp. 392 395
Jeong, H., Tombor, B., Albert, R., Oltvai, Z.N., and Barabasi, A.L.:
The large-scale organization of metabolic networks, Nature, 2000,
407, pp. 651 654
Albert, R., Jeong, H., and Barabasi, A.L.: Error and attack tolerance
of complex networks, Nature, 2000, 406, pp. 378 382
Han, J.D., Bertin, N., Hao, T., Goldberg, D.S., Berriz, G.F., Zhang, L.V.,
Dupuy, D., Walhout, A.J., Cusick, M.E., Roth, F.P., and Vidal, M.:
Evidence for dynamically organized modularity in the yeast protein
protein interaction network, Nature, 2004, 430, pp. 8893
Reichlin, S.: Neuroendocrine consequences of systemic inammation in (Committee on Military Nutrition Research (Ed.):
Military strategies for sustainment of nutrition immune function in
the eld (National Academy Press, Washington DC, 1999), p. 395
Besedovsky, H.O., del Rey, A., Klusman, I., Furukawa, H., Monge
Arditi, G., and Kabiersch, A.: Cytokines as modulators of the
hypothalamuspituitaryadrenal axis, J. Steroid. Biochem. Mol.
Biol., 1991, 40, pp. 613 618
59

32 Cutolo, M., Sulli, A., Pizzorni, C., Craviotto, C., and Straub, R.H.:
Hypothalamicpituitary adrenocortical and gonadal functions
in rheumatoid arthritis, Ann. NY Acad. Sci., 2003, 992
pp. 107 117
33 Pearce, G.J., and Chikanza, I.C.: Targeting tumour necrosis factor in
the treatment of rheumatoid arthritis, BioDrugs, 2001, 15,
pp. 139 149
34 Imura, H., Fukata, J.U., and Mori, T.: Cytokines and endocrine
function an interaction between the immune and neuroendocine
systems, Clin. Endocrinol., 1991, 35, pp. 107 115
35 De Simoni, M.G., Sironi, M., De Luigi, A., Manfredi, A., Mantovani,
A., and Chezzi, P.: Intracerebroventricular injection of interleukin 1
induces high circulation levels of interleukin 6, J. Exp. Med., 1990,
171, pp. 17731778
36 Chrousos, G.P.: The hypothalamicpituitaryadrenal axis and
immune-mediated inammation, N. Engl. J. Med., 1995, 332,
pp. 13511362
37 Perlstein, R.S., Whitnall, M.H., Abrams, J.S., Mougey, E.H., and
Neta, R.: Synergistic roles of interleukin 6, interleukin 1, and
tumor necrosis factor in the adrenocorticotropin response to
bacterial lipopolysaccharide in vivo, Endocrinology, 1993, 132,
pp. 946 952
38 Wilder, R.L., Grifths, M.M., Cannon, G.W., Caspi, R., and
Remmers, E.F.: Susceptibility to autoimmune disease and drug
addiction in inbred rats. Are there mechanistic factors in common

60

39

40
41
42

43
44

45

related to abnormalities in hypothalamic pituitary adrenal axis and

stress response function?, Ann. NY Acad. Sci., 2000, 917,
pp. 784 796
Chikanza, I.C., Petrou, P., Kingsley, G., Chrousos, G., and Panayi,
G.S.: Defective hypothalamic response to immune and
inammatory stimuli in patients with rheumatoid arthritis, Arthritis
Rheum., 1992, 35, pp. 12811288
Mastorakos, G., and Ilias, I.: Relationship between interleukin-6
(IL-6) and hypothalamicpituitary adrenal axis hormones in
rheumatoid arthritis, Z. Rheumatol., 2000, 59, pp. 7579
Dunn, A.J.: Interleukin-6 access to the axis, Endocrinology, 2004,
145, pp. 3578 3579
Li, S., Lu, A.P., Li, B., and Wang, Y.Y.: Circadian rhythms on
hypothalamicpituitary adrenal axis hormones and cytokines of
collagen induced arthritis in rats, J. Autoimmun., 2004, 22,
pp. 277 285
Petrovsky, N., and Harrison, L.C.: The chronobiology of human
cytokine production, Int. Rev. Immunol., 1998, 16, pp. 635649
Miyazawa, K., Mori, A., and Okudaira, H.: Regulation of
interleukin-1beta-induced interleukin-6 gene expression in human
broblastlike synoviocytes by glucocorticoids, J. Biochem. (Tokyo),
1998, 124, pp. 11301137
Cai, D.F., Shen, Z.Y., and Chen, X.H.: Effect of aconitine on content
of corticotropin-releasing hormone in hypothalamus of rats,
Zhongguo Zhong Xi Yi Jie He Za Zhi, 1996, 16, pp. 544 546

IET Syst. Biol., Vol. 1, No. 1, January 2007