Professional Documents
Culture Documents
PII: S1044-579X(18)30027-0
DOI: https://doi.org/10.1016/j.semcancer.2018.09.005
Reference: YSCBI 1513
Please cite this article as: Tan W, Zhong Z, Carney RP, Men Y, Li J, Pan T, Wang Y,
Deciphering the metabolic role of AMPK in cancer multi-drug resistance, Seminars in
Cancer Biology (2018), https://doi.org/10.1016/j.semcancer.2018.09.005
This is a PDF file of an unedited manuscript that has been accepted for publication.
As a service to our customers we are providing this early version of the manuscript.
The manuscript will undergo copyediting, typesetting, and review of the resulting proof
before it is published in its final form. Please note that during the production process
errors may be discovered which could affect the content, and all legal disclaimers that
apply to the journal pertain.
Deciphering the Metabolic Role of AMPK in Cancer Multi-Drug Resistance
Wen Tana, b#, Zhangfeng Zhongc, d#, Randy P. Carneye, Yongfan Menb, Jiannan Lib, Tingrui
T
R IP
a.
School of Pharmacy, Lanzhou University, Lanzhou, Gansu province 730000,
SC
China
b.
Micro-Nano Innovations (MiNI) Laboratory, Biomedical Engineering, University
d.
Institute of Chinese Medical Sciences, State Key Laboratory of Quality Research
e.
Department of Biomedical Engineering, University of California Davis, Davis,
TE
*Corresponding author
Prof. Yitao Wang, the State Key Laboratory of Quality Research in Chinese Medicine,
A
Institute of Chinese Medical Sciences, University of Macau, Macau SAR 999078, China.
Email: ytwang@umac.mo
1
Abstract
T
whereby some cells rapidly adapt to the tumor microenvironment via a myriad of
IP
heterogeneous metabolic activities. Despite being a major impediment to treatment, there is a
R
silver lining: control over metabolic regulation could be an effective approach to overcome or
SC
correct resistance pathways. In this critical review, we comprehensively and carefully curated
and analyzed large networks of previously identified proteins associated with metabolic
U
adaptation in MDR. We employed data and text mining to study and categorize more than
N
600 studies in PubMed, with particular focus on AMPK, a central and fundamental modulator
A
in the energy metabolism network that has been specifically implicated in cancer MDR
pathways. We have identified one protein set of metabolic adaptations with 137 members
M
closely related to cancer MDR processes, and a second protein set with 165 members derived
D
from AMPK-based networks, with 28 proteins found at the intersection between the two sets.
TE
Furthermore, according to genomics analysis of the cancer genome atlas (TCGA) provisional
data, the highest alteration frequency (80.0%) of the genes encoding the intersected proteins
EP
(28 proteins), ranked three cancer types with quite remarkable significance across 166 studies.
The hierarchical relationships of the entire identified gene and protein networks indicate
CC
and depict the metabolic roles of AMPK and demonstrate the potential of metabolic control
A
2
1. Introduction
Multi-drug resistance (MDR) is one of the major obstacles in chemotherapy and occurs in
T
final failure. The magnitude of MDR incidence is frightening, and the impact of MDR on
IP
cancer mortality is significant and worldwide. Overcoming MDR has been regarded as one of
the main promises of the era of personalized medicine, and different strategies have been
R
conducted to reduce clinical burden [1, 2]. First, it is important to understand that MDR is an
SC
evolving paradigm, featuring a series of mechanisms including drug-related factors (drug
transport and metabolism, alterations in drug targets) and cell response-related factors (DNA
U
damage repair, intrinsic adaptive responses promoting survival and resistance, a wider
N
chemical view of the tumor microenvironment) [3, 4]. Despite the heterogenous character of
A
MDR, most of the focus on intervention has involved improving drug transport, for example
M
by developing new delivery systems to overcome low intracellular drug concentrations that
occur via active drug expulsion by MDR cancer cells. Yet these methods treat the symptoms
D
of MDR, not the underlying causes. More recently, a novel approach of reprogramming
TE
Tumor tissue exhibits vast cellular heterogeneity, dictated by a dynamic and complex
spatial and chemical system known as the tumor microenvironment. Various regions of the
A
chemical gradients of oxygen, sugar, and other biomolecules. Acquisition of cancer resistance
downregulate or upregulate biomolecules acting in key energy metabolism pathways [6]. For
3
example, cancer cells have the capacity to adapt to hypoxic and hypo-nutrient conditions,
such conditions by active glycolysis and autophagy, altered lactate production, and altered
K562 cells or NCI-H460 cells, their MDR counterparts (K562Dox cells or NCI-H460/R) have
T
increasing rates of glycolysis and methylation capacity, decreasing rates of pentose phosphate
IP
pathway (PPP) and oxidative phosphorylation (OXPHOS), and altered glutathione
metabolism [9]. In highly resistant multiple myeloma cells (AMO-BTZ and AMO-CFZ), a
R
large group (>600) of aberrantly-expressed proteins were identified and further found to be
SC
involved with metabolic control and redox homeostasis pathways, including ones related to
the respiratory chain, metabolite generation, glycolysis, and metabolism of amino acids,
U
lipids, and nucleic acids. It is clear from previous reports that there is a deep relationship
N
between metabolic reprogramming and cancer resistance [10].
A
Given that metabolic reprogramming majorly drives MDR progression, it is likely that
D
metabolic control could be an effective approach to overcome MDR. This potential has been
TE
confirmed by the novel perspective of omics-related findings in recent years, which have
helped to identify personalized anticancer treatments and new therapeutic targets [11-13].
EP
been attempted. For example, metabolic adaptation in cancer stem cells is regarded as a major
CC
potential cause of chemotherapy failure [14]. In one study, following chemotherapy failure in
blocking of the related mitochondrial respiratory chain was found to overcome MDR [15].
model has also been confirmed to be a viable approach to treat MDR [16]. Regulation of
energy sources in the tumor microenvironment has been shown to restore chemo-sensitivity
4
[17, 18]. In that approach, bioenergetic modifiers work to increase sensitivity to conventional
known as the “Warburg effect” [19]. Metformin, a first-line drug for type 2 diabetes, which
metformin with 2-deoxylucose further aggravated metabolic stress on resistant cancer cells
T
and induced a series of cell cycle arrest and apoptosis, increasing doxorubicin accumulation
IP
[21, 22]. It is clear that metabolic control is a high-potential therapeutic strategy capable of
proactive management of energy metabolism in cancer cells, bringing along a new path to
R
overcome MDR for lower morbidity and mortality rates.
SC
In Section 2, we review the major players associated with the AMPK-mediated energy
U
metabolism network, focusing on key signaling pathways related to MDR, while highlighting
N
successful metabolic targets that have been reported in the literature. In Section 3, we
A
perform critical text-mining on those reports to reveal a highly interconnected and complex
regulatory network that largely dictates MDR in cancer. It is our hope that this global view
M
will aid researchers in developing more effective therapeutic interventions targeting metabolic
D
The AMPK enzyme is regarded as a sensor of energy and nutrients in cells. It is comprised
EP
regulatory subunits 𝛽 (5'-AMP-activated protein kinase subunit beta-1 and 2, PRKAB1 and
CC
PRKAB2), and three regulatory subunits 𝛾 (5'-AMP-activated protein kinase subunit gamma-
and phosphorylated-AMPK was used as an independent prognostic marker for patients with
resected gastric cancer, where it significantly correlated with longer survival (61.0% in 621
patients) [23]. AMPK activation mediates metabolic reprogramming in resistant cancer cells
through promoting both the Warburg effect and also mitochondrial biogenesis. It also
5
regulates the self-renewal ability of cancer stem cells and induces autophagy, a double punch
for chemotherapy resistance [24]. A complex network of energy metabolism surrounds the
central AMPK enzyme, illustrating both its ubiquitous role in energy balance and also the
substantial amount of crosstalk between AMPK and other signaling pathways. These
signaling pathways involve (i) some pro-drugs converting to AMP analogs in cells, (ii) some
compounds inhibiting mitochondrial function and increasing AMP/ADP, or (iii) some direct
T
activators of AMPK [25, 26]. Among them, the most concentrated field is glycolysis,
IP
mitochondrial-related, and fatty acid synthesis (FASN), but also include glutamine
R
SC
2.1 An Indispensable Glycolysis and a Pentose Phosphate Pathway (PPP)
coordinator
U
Glycolysis is an indispensable and routine way cancer cells use to extract energy. The
N
pentose phosphate pathway (PPP) coordinates glycolysis, biosynthesis, and redox regulation
A
through an ingenious approach, producing ribose-5-phosphate (Ru-5-P), NADPH, and
M
cancers with the “Warburg effect”. As the most pivotal player in the “Warburg effect”, HK2
D
drives the prominent phenotype (high aerobic glycolysis) and helps immortalize cancer cells
TE
[27, 28]. For human breast carcinoma, HK2 activity and intracellular distribution indicated
invasiveness and aggressiveness, among other major prognosis markers [29]. In glioblastoma
EP
or PFKP) provides an alternative way to survival of cancer cells even under hypoxia
in response to hypoxia, thereby inhibiting PFK activity and redirecting glucose metabolism to
PPP. It was observed that blocking the O-GlcNAcylation of PFK at Ser529 site leads to a
reduction of cancer cell proliferation in vitro and impaired tumor formation in vivo [31].
6
and functional mutation can also switch the flux of glucose towards PPP for cancer cell
survival and metastasis. For example, Snail suppresses the platelet isoform of PFK-1 (PFKP)
and turns to PPP for cancer cell survival under metabolic stress and increasing metastasis in
vivo [32]. Functional significance of selective PFKP mutations also contribute to metabolic
states in certain cancers [33]. The up-regulated predominant isoform of PFK-1, PFKP, is
found in human clear cell renal cell carcinoma (RCC) and has been associated with p53
T
suppression and malignant proliferation [22]. PFKP is mainly regulated by 6-phosphofructo-
IP
2-kinase/fructose-2,6- bisphosphatase 3 (PFKFB3, or PFK-2), the latter is regulated by
R
induced mitophagy in cancer cells is achieved by AMPK, and PFKFB3 mediates glycolysis
SC
for survival [34]. 6-phosphofructo-2-kinase/fructose- 2,6-biphosphatase 4 (PFKFB4) is an
isoform of PFKFB3 and exerts balancing glycolytic activity and antioxidant production for a
(PEP) to pyruvate. Inhibition of PKM2 might be mediated by AMPK activation, and applied
M
to breast cancer cells, resulted in apoptosis due to low levels of glucose or amino acid [36].
D
PDH at its Ser294 site, which inactivates the PDC and leads to lactate production instead of
EP
the conversion of pyruvate to acetyl-CoA typically catalyzed by PDC. There is some evidence
of an indirect relationship between PDK and AMPK, since silencing pyruvate dehydrogenase
CC
kinase isoform 4 (PDK4) attenuates the phosphorylation level of AMPK induced by casein
kinase 2 (CK2), a result that could be produced in vitro and in vivo [37]. In low-nutrient
A
conditions in HeLa cells, PDK activation inhibits PDH phosphorylation and eventually drives
survival of cancer cells in a AMPK-dependent manner [38]. Thus, there is evidence that in
certain situations, PDK and AMPK exhibit a synchronized pace. Conversely, PDK4 was
observed in cancer cells to bypasses AMPK activation and subsequently activate mTOR
complex 1 (mTORC1) for aerobic glycolysis and tumorigenesis [39]. Lactate dehydrogenase
7
A (LDHA) prefers to promote lactate production, and exhibits a similar pattern to AMPK
LDHA and AMPK were consistent with TNBC stage, distant metastasis, Ki67 status, and
survival outcome of patients [40]. Meanwhile, LDHA may be linked with AMPK activation
to support a glycolytic phenotype through the repression of mTORC1 inhibition and the
subsequent increase of HIF-1𝛼, which could bind to specific sites on LDHA promoter in
T
cancer cells [41]. Glucose-6-phosphate dehydrogenase (G6PDH) catalyzes the first
IP
committed step of PPP, converting glucose-6-phosphate to 6-phosphogluconolactone and
R
inhibited by AMPK activation in arachidonic acid-treated hepatocytes [42]. In cancer cells, 6-
SC
phosphogluconate dehydrogenase (6PGDH), the third enzyme in PPP, stimulates the
U
disruption of the active LKB1 complex [43]. NADPH production from PPP is impaired when
N
cancer cells face energy stress, for instance, during glucose limitation, anchorage-independent
A
growth, or solid formation in vivo. AMPK activation occurs to decrease NADPH
M
consumption in fatty acid synthesis and increase its production in fatty acid oxidation to
Citrate synthase (CS), the first enzyme of the tricarboxylic acid (TCA) cycle, is highly
EP
expressed in malignant ovarian tumors and ovarian cancer cells. CS knockdown enhances
sensitivity to cisplatin in SKOV3 and A2780 cells and decreases phosphorylated AMPK [45].
CC
proliferation, invasion, and migration, and promotes apoptosis in ovarian cancer cell lines
A
SKOV3 and A2780, accompanied by a reduction of phosphate AMPKα [46]. Notably, a loss
of another tumor suppressor, fumarase (FH), induces fumarate accumulation that activates
AMPK, which in turn confers protection from apoptosis in hereditary leiomyomatosis and
renal cell cancer syndrome (HLRCC) [47]. Malate dehydrogenase 2 (MDH2) inhibition
induces AMPK activation in human colorectal cancer HCT116 cells [48]. This mechanism is
8
embodied in a novel MDH2 inhibitor that exhibits significant growth inhibition of HCT116
cells in vitro and tumor growth suppression of HCT116 xenograft mouse model in vivo [49].
AMPK causes ACC inhibition. Besides ACC inhibition, ATP-citrate lyase (ACLY) is another
T
promotes energetic stress and AMPK activation, which supports androgen receptor (AR) level
IP
maintaining, inhibition of proliferation, and apoptosis induction. An ACLY-AMPK-AR
R
feedback loop might be applied to sensitize CRPC cells to AR antagonism for preventing
SC
inappropriate reactivation of AR after androgen deprivation therapy [50]. In colon cancer,
phosphate AMPK levels correlated with cellular ROS levels in vitro and in vivo, with lower
U
basal levels found to be critical for the anticancer effect of ACLY depletion [51]. From a co-
N
immunoprecipitation assay, an in vitro glutathione S-transferase (GST) pulldown assay, and
A
an in vitro kinase assay, the inhibition of ACLY on AMPK activation was confirmed to be
M
due to a physical interaction with the catalytic subunit 𝛼2 [52]. Stearoyl-CoA desaturase 1
(SCD1) is a rate-limiting enzyme that catalyzes the conversion of saturated fatty acids (SFA)
D
into monounsaturated fatty acids (MUFA), the main building blocks for phospholipids,
TE
ACC, accompanied by in vitro invasion suppression and tumor growth inhibition [53].
Glutamine is an important source of carbon and nitrogen for malignant cancer cells [54].
A
and anchorage-independent colony formation of cancer cells and induce autophagy through
9
by glutamic-oxaloacetic transaminase 1 (GOT1) to produce glutamate. In addition to
glutamate metabolism, pH regulation and serine synthesis are also viable targets. Solute
T
2.5 AMPK-mediated approaches overcoming MDR in experimental research
IP
There are some clinical drugs, small molecular inhibitors and bioactive compounds derived
R
from natural products exerting MDR improvement by means of AMPK-mediated metabolic
SC
modulation. Here the Table 1 illustrates these compounds targeting AMPK-mediated
pathways in MDR with in vitro or in vivo investigations briefly and concludes the effects and
U
the potential mechanisms. In majority of the involved pathways, AMPK is activated and
N
amplified by the downstream effectors, such as mTOR inhibition. Besides, multidrug
A
resistance protein 1 (MDR1) inhibition at gene and/or protein level, down-regulations on
M
multidrug resistance-associated protein 1 (MRP1), NF-𝜅B signaling suppression, are the most
common pathways against MDR. Autophagic or apoptotic cell death, the enhanced
D
cytotoxicity and proliferation inhibition, the cell cycle arrest, and metastasis or invasion
TE
attenuation are found. Among the energy metabolism network, only a limited amount of
proteins (e.g. ACACA [55], ACLY [56], HK2 [57] and LDHA [58] in Table 1) are focused
EP
with further study for in-deep research, and majority of them have not been explored yet,
especially on the specific metabolic network of AMPK-mediated. It might also be noted that
CC
the role of AMPK might be context dependent and the exact mechanism in specific cancer
type or tumor microenvironment should be viewed with some caution. Since many of the
A
compounds in Table 1 have multiple actions, in our present study we focus on the specific
control.
10
R I
SC
Table 1 List of compounds targeting AMPK-mediated pathways in MDR
U
Compo Classific Refere
N
Resistant
und ation/Ta Pathway Effect nce
Model
A
Name rget (Year)
M
In vitro
Cisplatin-
ED Enhancing Inducing autophagic
resistant [59]
phosphorylation of cell death and
human oral (2017)
AMPK apoptotic death
cavity
PT
cancer cell
Chemo-
Reducing Increasing oxaliplatin
E
Resver preventive
In vitro transcriptional toxicity in
atrol phytochem
CC
(2015)
cancer cell Down-regulating oxaliplatin in
HCT116/L- MDR1 expression by HCT116 cells;
OHP inhibiting degradation Increasing
and phosphorylation accumulation of
11
R I
SC
of I𝜅B𝛼 and p65 Rh123 in resistant
U
translocation, and cells;
N
through CRE
transcriptional
A
activity decrease
M
mediated by
ED phosphorylation of
AMPK;
The reduction in
In vitro Reducing phosphorylation of
PT
activator (2016)
gastric Reducing mdr1 gene resistance in 5-FU-
CC
12
R I
SC
CRPC p70S6K-Mcl-1 axis dependent pro-
U
mouse and induces ULK-1 survival autophagy
N
model signaling
In vitro
A
Having a synergistic
Human Activating AMPK and
anti-proliferation
M
hepatocellul inhibiting mTOR
effect with 5-Fu;
ar phosphorylation;
ED Promoting apoptosis
carcinoma Decreasing
and inducing cell
Bel-7402/5- expressions of HIF- [62]
G0/G1 cycle arrest;
Anti- Fu cell; 1𝛼, MDR1 and (2014)
PT
Resensitizing MCF-
Metfor diabetes Acquired MRP1; [63]
7/5-Fu cell and
min biguanide drug- Inhibiting MDR1 (2014)
E
innate drug-resistant
drug resistant expression by [64]
CC
MDA-MB-231 cell
MCF-7/5- blocking MDR1 gene (2010)
and promoting
Fu cell and transcription;
apoptosis;
A
13
R I
SC
Human Reducing NF-𝜅B Increasing
U
breast activity and I𝜅B adriamycin toxicity
N
cancer degradation; in MCF-7/adr cell;
adriamycin Inhibiting
A
resistant phosphorylation of
M
cell line CREB;
ED(MCF-
7/adr);
Convergent activation
of pathways Imposing a great
PT
In vitro
underlying tumor selective pressure for
Human
microenvironment cellular states;
E
breast
interactions leads to a Triggering a [65]
CC
cancer
notable transcriptome (2014)
metformin-
downregulation of the reprogramming
resistant
A
14
R I
SC
stability and pro-
U
autophagic features;
N
Increasing the
potential of metastatic
A
dissemination by
M
amplifying the
ED number of pro-
migratory and
stemness inputs via
the activation of a
PT
significant number of
proteases and EMT
E
regulators;
CC
15
R I
SC
PLC/PRF/5, Enhancing apoptosis
U
HuH6, induced by sorafenib,
N
HLE, HLF, accompanied by
Hep3B) activation of p38
A
MAPK and JNK,
M
upregulation and
ED translocation of Bax,
and activation of
caspase-3;
Active
PT
sesquiterpe Regulating
ne isolated In vitro Activation AMPK; mitochondrial
E
diene Curcuma n-resistant ACLY and phosphate reducing ATP level; (2016)
wenyujin MCF-7 cell GSK-3𝛽; Inducing apoptotic
A
16
R I
SC
melian d Paclitaxel- transcription of
U
odiol derivative resistant MDR1;
N
of A549 cell Regulating
Poncirus PI3K/mTOR
A
trifoliate signaling pathway;
M
In vitro
EDHuman
colon Inducing necrosis
cancer and apoptosis when
HCT-116 Activating AMPK and combined with
PT
cerami
permeable cancer mTORC1; doxorubicin-induced [67]
CC
de
ceramide PANC-1 Down-regulating Bcl- apoptosis in vitro; (2010)
cell and 2/HIF-1𝛼; Enhancing anti-
A
17
I
R
SC
A2780 cell
U
and
N
CaOV3,
human
A
breast
M
cancer
MCF-7 cell,
ED non-
Hodgkin’s
lymphoma
PT
KARPAS-
422 cell and
E
Hodgkin’s
CC
lymphoma
L428 cell,
A
melanoma
cancer
WM-115
cell;
In vivo
18
R I
SC
HCT-116 or
U
A2780
N
xenografts
In vitro
A
Activating AMPK and Promoting AMPK-
Bortezomib
Bortez Proteasom promoting ULK1/2 to dependent [68]
M
-resistant
omib e inhibitor form a complex with autophagosome (2014)
myeloma
ED ATG13; formation;
cell
Increasing cell
In vitro response to
PT
molecular (2014)
WM266 conditioned medium
CC
19
R I
SC
cell, Mz- Enhancing sensitivity
U
ChA-1 cell, of GBC-SD cell to
N
QBC939 gemcitabine and 5-
cell Fu;
A
Down-regulating P-gp
M
In vitro expression;
Bioactive
Human Activating AMPK;
compound ED breast Inhibiting NF-𝜅B Reversing P-gp-
Mollug from roots [71]
cancer signaling pathway and mediated multidrug
in of Rubica (2013)
MCF- COX-2 expression resistance
PT
cordifolia
7/Adriamyc and attenuating CRE
L.
in cell transcriptional
E
activity;
CC
resistant
Sorafen B-Raf kinase activity; proliferation; [72](20
targeting Ras-NIH
ib Abolishing both ERK Inducing autophagy 13)
drug 3T3/Mdr and MER and apoptosis;
cell phosphorylation;
20
R I
SC
Inhibiting LKB1 Reversing the
U
expression and resistance to
N
inducing a sustained paclitaxel;
activation of AMPK
A
at a concentration of 5
M
𝜇M;
ED Inducing decreases of
mTOR
phosphorylation at
Ser-2448, and
PT
phosphorylation of
p70S6K and 4E-BP1;
E
Negatively regulating
CC
P-gp function;
21
R I
SC
NIH 3T3 Delay Raf activation;
U
cell
N
A
M
Inhibiting MDR1
ED expression and
suppressing MDR1
In vitro
mRNA and MDR1
PT
Bioactive Human
promoter activity;
compound breast
Puerari Increasing drug Reversing cancer [55]
E
from cancer
n accumulation of drug resistance; (2010)
Pueraria MDR cell
CC
doxorubicin;
lobata line (MCF-
Inhibiting NF-𝜅B
7/adr cell)
A
22
R I
SC
Inducing
U
phosphorylation of
N
ACC and GSK-3𝛽;
Decreasing CREB
A
phosphorylation;
M
ED
EPT
CC
A
23
3. Analysis of Mined Protein Alterations Related to MDR and Cancer Metabolism
In the present study, we employed text and data mining to identify and map all reported
were searched in PubMed with the terms (“Drug Resistance, Neoplasm [MeSH])” and
T
(“Energy Metabolism [MeSH]”). By 18 September 2017, a total of 636 publications were
IP
retrieved and reviewed. First, abstracts of the selected publications were used to screen for
R
pharmacological studies associated with the alterations of key proteins known to be involved
SC
in energy metabolic pathways. Next, to ensure the applicability of each study to our present
review, the full reports of the selected publications were reviewed rigorously. Our inclusion
U
criteria were proteins that (i) exhibited definite alterations in resistant cancer cells or
N
transformed cancer cells with the MDR phenotype and (ii) were reported in experimental data
A
across multiple pharmacological approaches. Studies performed with these criteria in cancer
M
stem cells were also included, given their well-known MDR characteristics [74-76]. Studies
involving MDR intervention induced by novel drug delivery systems were ruled out. Those
D
criteria identified one key protein subset composed of 137 metabolic adaptations with many
TE
related to reported cancer MDR processes. A detailed list of the altered proteins is shown in
Supplement 1.
EP
Primary among the 137 altered MDR-related proteins identified are glucose transportation
and metabolism related proteins (GLUT1, 2, 3, 4, 10, HK1, HK2, PFKFB2, PFKFB3, PGM1,
CC
PGM2), aerobic glycolysis related proteins (PFKP, ALDOA, PKM2, LDHA and B, GAPDH,
MLXIP, PGAM1), mitochondrial and energy metabolism related proteins (ATP5A1, ATP5B,
A
ATP5D, CYC1, DNM1L, IDH1, LKB1, MCU, MDH, ME1, MT-CO1, mTOR, NRF1,
PDHA1, PDK1, PDK2, PDK3, PDK4, PEPCK, PPARGC1A, PPIF, PRDX1, PYGB, SDHA,
SDHB, SIRT1, TFAM, TKT, UCP1, UCP2, UMPS, UQCRC1, UQCRC2), as well as lipid
metabolism process related proteins (FASN, ACACA, ACLY, ACSF2, ACSF3, PTEN).
24
Others include members of the aldo-keto reductase family, AKR1B10 (Aldose reductases),
that regulate lipid peroxidation and the metabolism of corticosteroids, biogenic amines, and
T
NADP/NADPH related proteins (APOA1BP, BLVRA, NDUFB8, NOX2, NQO1),
IP
glutathione metabolism related enzymes (GCS, GGT, GOT1, GPX1, GSTM1, GSTM4,
MGST1, MGST3), and proteins responsible for glucose uptake regulation (AKT1 and 2) were
R
also selected. Still others include: the ESRRA/PGC1alpha complex, a regulator of energy
SC
metabolism; G6PD, the rate-limiting enzyme of pentose-phosphate pathway; GLS1 and
GLUL, glutamine metabolism related enzymes; and finally, HIF-1A, the hypoxia
U
conditioning protein closely related to glucose transporters, glycolytic enzymes, and
N
metabolic adaptation. The diversity of the proteins significantly associated with metabolic
A
adaptation in cancer MDR clearly indicate a highly complex and dynamic regulatory network.
M
Of the 137 altered proteins reported in pharmacological studies to be associated with the
TE
metabolic reprogramming adaptation of resistant cancer cells (as shown in Supplement 1),
and especially weighing the different roles of AMPK in these metabolic processes (as
EP
cells has also influenced our selection criteria [5, 77, 78]. For instance, in glucose
CC
SLC16A) were chosen as the initial objects to further explore deeper correlations. While in
(IDH1), IDH2, succinate dehydrogenase complex (A, B, C, D), FH, MDH, were selected as
the initial data. For fatty acid synthesis, ACLY, ACC, HMGCR, FASN and SCD were
25
selected. Finally, GLS, GLUD, GOT1, SLC9A1, CA9, G6PD and PHGDH were included to
To assess and integrate protein and protein interaction (PPI) networks we utilized the
networks identified from many sources, including text mining (networks extracted from the
T
biophysical assays), other curated databases, co-expression (protein-encoding genes co-
IP
expressed in the same or in different species), neighborhood analysis (repeatedly occurring
R
spatially close networks of protein-encoding genes), gene fusion (gene fusion events per
SC
species for protein-encoding genes), and co-occurrence (the presence or absence information
on linked proteins across species) [79, 80]. In our present study, to comprehensively and
U
efficiently predict direct or indirect associations, PPI networks were characterized using
N
systematic co-expression data, database annotations, and automated text mining of the
A
scientific literature for each metabolic process. By including statistical and/or semantic links
between proteins derived from Medline abstracts and full-text articles, a widely
M
The four panels of Figure 1 illustrate the specific interactions of AMPK (represented as a
TE
grouped circle comprised of seven key proteins) in each identified metabolic process, i.e.
respectively. The minimum required interaction score was set as medium confidence (0.400)
for a broad and comprehensive analysis. The max number of interactions included was no
CC
more than 50 interactors. Networks were clustered using a Markov cluster algorithm (MCL)
[81]. Key proteins with interaction value greater than 0.500 in automated text mining were
A
selected. Of those, 56.75% (668 PPIs among 1177 PPIs) were related to glycolysis (76
proteins), 41.97% (410 PPIs among 977 PPIs) to mitochondrial-related (59 proteins), 44.86%
(388 PPIs among 865 PPIs) to fatty acid synthesis (54 proteins), and 54.63% (336 PPIs
among 615 PPIs) to glutamine or related processes (55 proteins). Cytoscape, a software
platform that builds 2D depictions of molecular interaction networks [82], was applied to
26
visualize the identified key proteins and interactions as depicted in Fig 1. Nodes represent
proteins and edges (the lines between two nodes) represent protein-proteins associations. The
size of the nodes indicates the level of functional associations between proteins according to
number of database annotations. Node color was derived from co-expression data. Edge color
represents the variance of interactions parsed across primary databases during automated text
mining data. To better emphasize the main trends of the entire network in each metabolic
T
process, multiple edges between two nodes were bundled together, generating overlapping
IP
lines.
R
The hierarchical relationships of the entire network can be easily analyzed by considering
SC
together the main trends, node size, and node/edge color. Accordingly, rather than simple
causal relationships, the broadest possible correlations among proteins is displayed, therefore
U
this depiction indicates a full-scale potential PPI network bundled from the perspective of
N
AMPK-mediated metabolic control of MDR. For example, as shown in Fig 1A, there are five
A
groups of highlighted protein with intensive interactions in glycolysis, namely the first group
(HK1, SLC2A2, PC, LDHA, PFKM, PFKL, PGM1, PGAM1, GMPS, MDH2), the second
M
group (TALDO1, HK3, ENO4, ACACB, FH, ENO3, PFKFB4, PKM, PGK2, PGK1,
D
ACACA), the third group (TKT, LDHAL6B, MDH1, PFKP, LDHAL6A, GPI, SLC2A4), the
TE
fourth group (PFKFB3, HK2, FBP2, PDHA2, FBP1, PDHA1, ENO2, ENO1, ALDOC,
LDHB) and the fifth group (LDHC, ALDOA, ALDOB, TPI1, ME2, ME3, ME1, PKLR,
EP
PFKFB2), without priorities in sorting. Similarly, in Fig 1B, six groups are found with
ME3, ME2, SUCLG2, PDHX, CS, IDH3A, DPYD), the second group (GMPS, ATP5C1,
PCK1, DLAT, IDH3G, ACO1, IDH3B, NDUFS2, NDUFS3), the third group (SDHAF2,
A
SDHD, TP53, FH, GOT2, ACLY, PDHB, PC, PDHA2, PDHA1, DHTKD1, BCKDHA,
ACACB, BCKDHB, ACACA), the fourth group (IDH1, GOT1, IDH2, MDH1, SUCLA2,
DLD, PDK2, PDK1, MTOR), the fifth group (SDHC, OGDH, SDHA, UQCRFS1, SUCLG1,
OGDHL, SDHB, MDH2, ATP5B, FXN), the sixth group (HSPD1, ACO2, VDAC1, CYCS,
RPTOR). There are several groups of highlighted protein with intensive interactions in fatty
27
acid synthesis related processes (Fig1C) and in glutamine-related processes (Fig1D).
Although most of proteins in these potential PPI networks have been confirmed by practical
experiments, only a few of them (such as PKM [83], PDK [18], etc.) are focused on MDR
improvement by metabolic control. Therefore, more proof and in-depth studies on this
In addition to the associations between the proteins comprising each metabolic process, we
T
analyzed the core proteins directly interacted with AMPK (including PRKAA1, PRKAA2,
IP
PRKAB1, PRKAB2, PRKAG1, PRKAG2, and PRKAG3). Heat maps representing the
R
computed text-mining scores are shown in Figure 2, similarly broken down into the four
SC
identified metabolic pathways. More proteins across each pathway were found to preferably
interact with the catalytic subunits PRKAA1 and PRKAA2. Only a few proteins prefer to
U
interact with PRKAG1 and PRKAG2, the two regulatory subunits. For the catalytic subunits
N
of AMPK, four proteins were found to participate in all the metabolic process presented in
A
our study: ACACA (Acetyl-CoA carboxylase alpha, ACACA), ACACB (Acetyl-CoA
Figure 3A, these directly linked proteins comprise a heterogeneous group with 22 members
TE
(ACACA, ACACB, ACLY, ASNS, CAB39, CAD, GOT2, HK2, HMGCR, LEP, MTOR,
PFKFB2, PFKFB4, PPARGC1A, RHEB, RPTOR, SLC2A4, STK11, TP53, TSC1, TSC2,
EP
ULK1). This grouping helps to visualize the central role of AMPK in the wider energy
metabolism network. The key proteins with the closest interactions with AMPK in each
CC
metabolic process (text-mining score greater than 0.500) were collected for further analysis,
revealing a large AMPK-based network with 165 proteins (Supplement 2) that could be
A
broken down into four groups: glycolysis (76 proteins) (Supplement 3), mitochondrial-
related (59 proteins) (Supplement 4), fatty acid metabolism (54 proteins) (Supplement 5),
of the complex relationships between these four groups, with the proportion of shared
proteins amongst the groups represented as percentage values. The metabolic processes of
28
mitochondrial-related and FASN share the maximum amount of communal proteins, with 14
proteins accounting for 8.5% (ACO1, ACO2, CS, DHTKD1, DLAT, DPYD, OGDH,
OGDHL, PDHX, SDHA, SDHB, SDHC, SUCLA2, and SUCLG2). 9 proteins are shared
amongst all four groups, accounting for 5.5% (ACACA, ACACB, ACLY, GMPS, MDH1,
MDH2, MTOR, RPTOR, and TSC2). Other notable associations include 5 proteins (3%)
shared between glycolysis and others (GPI, HIF1A, PGAM1, PHGDH, and PSAT1), 5
T
proteins (3%) shared between mitochondrial-related and others (CAB39, GOT1, GOT2,
IP
TP53, and ULK1), 4 proteins (2.4%) shared among glycolysis, FASN, and others
(LDHAL6A, LDHAL6B, LDHB, and LDHC), 3 proteins (1.8%) shared among glycolysis,
R
mitochondrial-related, and FASN (ME1, ME3, and PC), 2 proteins (1.2%) shared among
SC
glycolysis, mitochondrial-related, and others (PPARGC1A and STK11), and 2 proteins
(1.2%) shared among mitochondrial-related, FASN, and others (IDH1 and IDH2).
U
N
We further analyzed and identified the most fundamental intersection between AMPK-
A
related energy metabolism network and metabolic adaptation in resistant cancer cells in order
to decipher the metabolic role of AMPK in MDR. In Figure 4, the key proteins (165 proteins)
M
derived from AMPK-based network through data mining and the altered proteins (137
D
proteins) in metabolic adaptation are compared. 28 proteins were found at this intersection,
TE
accounting for 10.2% of the total: ACACA, ACLY, ALDOA, ATP5B, ENO1, FASN, G6PD,
GAPDH, GLUL, GOT1, HK1, HK2, IDH1, LDHA, LDHB, ME1, PDHA1, PDK1, PDK2,
EP
PFKFB2, PFKFB3, PFKP, PGAM1, PGM1, PPARGC1A, SDHA, SDHB, and TKT. These
can be divided into several groups, fatty acid synthesis related group (Acetyl-CoA alpha,
CC
29
platelet, PFKP; Phosphoglycerate mutase 1, PGAM1; Phosphoglucomutase 1, PGM1;
T
These identified core proteins with dual roles in AMPK-mediated metabolic regulation in
IP
MDR may help take AMPK-mediated metabolic regulation from lab bench to clinical
R
bedside. Actually, there have been extensive studies on some of the pivotal proteins analyzed
SC
above, and the various proportions of contribution to the final metabolic phenotype or MDR
development were also investigated continuously in recent years. Finally, prior to verifying
U
these extracted proteins from literature analysis and data mining, the achievements of
N
experimental researches on AMPK-mediated network in overcoming MDR should also be
A
reviewed carefully.
M
4. The Perspective of Metabolic Control from the Clinic and in Drug Discovery
research in the past few decades. Such approaches focus on intra-tumoral heterogeneity and
TE
the tumor microenvironment, in the context of cellular metabolic control [84]. Many studies
have applied this paradigm to drug discovery. In this section, we apply various gene networks
EP
to analyze 28 genes responsible for encoding the core proteins extracted from the AMPK-
mediated metabolic pathway in MDR. Mutation analysis of these genes in clinical data was
CC
conducted with the cBio portal method, an open access resource for interactive exploration of
cancer data sets [85, 86]. Greater than 30% of altered genes in these core 28 genes is criteria,
A
hence 57 studies were found according to the cancer genome atlas (TCGA) provisional data,
as shown in Figure 5A. With quite remarkable significance across 166 various studies, a high
alteration frequency (>80.0%) ranked three cancer types: glioma (100%) [87], breast cancer
(96.6%) [88] and melanoma (80.6%) [89]. Glioma data represents sequence analysis of 23
30
initial low-grade gliomas patients, and the recurrent tumors resected from these patients.
Breast cancer data are from a genetically-defined clonal population dynamics study using
patient-derived xenografts (PDXs). Melanoma data are from a genomic analysis of pre-
For each caner type, genetic alterations for each gene involved in AMPK-mediated
metabolic pathways of MDR, which were found in at least three studies, are visualized
T
according to their percent alteration frequency in Figure 5B. In glioma, the gene alteration
IP
frequencies indicate the critical role of IDH1 mutation in resistance progression. Of the 28
R
genes across subtypes of glioma, IDH1 has 100%, 77%, and 73% of altered frequencies in
SC
low-grade glioma, brain lower grade glioma, and anaplastic oligodendroglioma or anaplastic
U
than that of other core genes, which may distinguish it as the most important gene encoding
N
protein in metabolic control for glioma resistance treatment. Actually, IDH1 is an important
(α-KG) and carbon dioxide. Hot-spot mutations in IDH1 are extensively found in malignant
M
IDH1 can be used as a non-invasive imaging approach with high sensitivity and specificity
TE
before individualizing surgical strategies. Consistent with our findings regarding IDH1, the
high frequency of IDH1 mutations in malignant glioma are being studied in clinical trials, and
EP
mutant IDH inhibitors have been developed into a treatment [93]. Similarly, for breast cancer,
as shown in Figure 5C, the highest alteration frequency of genes in three independent studies
CC
are: for PDXs – PFKP (52%), PFKFB3 (38%), and SDHA (34%); for adenoid cystic
carcinoma – ATP5B (42%), PDK2 (25%), and FASN/ACLY/ACACA (17%); and for
A
invasive breast cancer – PFKFB2 (25%), GLUL (22%), and PDK2 (7%). Some of these have
been investigated with good potential, such as PFKP [94], PFKFB3 [34, 95], SDHA [96],
ATP5B [97], PDK2 [98], FASN [99-101], ACLY [102, 103], ACACA [104, 105], and GLUL
[106, 107]. In the context of AMPK-mediated metabolic control in MDR, the roles or
underlying mechanisms of the proteins identified above, including PFKP, PFKFB3, SDHA,
31
ATP5B, PDK2, FASN, ACLY, ACACA, PFKFB2, and GLUL, need to be further
background makes metabolic modulation much more sophisticated than it was. And we have
to define the precise mechanisms with more thorough and cautious considerations before
melanoma, the most remarkable genes are FASN (19%, 10% and 5% for melanoma, skin
T
cutaneous melanoma, and desmoplastic melanoma respectively), ACACA (14% and 8% for
IP
melanoma and skin cutaneous melanoma), GLUL (16% and 5% for melanoma and
desmoplastic melanoma), ME1 (25% and 9% for desmoplastic melanoma and skin cutaneous
R
melanoma), PPARGC1A (25%, 14% and 7% for desmoplastic melanoma, melanoma, and
SC
skin cutaneous melanoma respectively), and PFKFB2 (14% and 7% for melanoma and skin
cutaneous melanoma), which might encode the most potential proteins in AMPK-mediated
U
metabolic pathway of melanoma MDR and be worthy of further study in the future [108-112].
N
For example, elevated expression of FASN is found to be associated with melanoma
A
progression [109], and a FASN inhibitor derived from Herba Epimedii exerts anti-melanoma
M
activities [110]. Besides the perspective of these extracted genes in data analysis of clinical
researches, there are some metabolic modifiers being studied on these proteins in drug
D
(targeting on PDK2), lonidamine (targeting on HK1), C75 (targeting FASN) and 6-NBDG
(targeting on HK1) [113, 114]. For example, DCA is an acid analogue of acetic acid
EP
inhibiting PDK2 with some ongoing clinical trials for safety and effectiveness
control in MDR, more and more metabolic modifiers are expected to be discovered, and their
A
corresponding metabolic approaches will come to the fore in future drug discovery.
Conclusion
therapy. Based on our critical review of studies on metabolic modifiers that confer attenuation
32
in cancer resistance, and our subsequent presentation of the deep, central metabolic role of
AMPK, we posit that methods which interfere with AMPK metabolism would be compelling.
Through a stringent text mining approach, our present study provides a fully described
AMPK-mediated metabolic network in MDR represented by a vivid PPI figure, along with
analysis of the key proteins with the closest interactions with AMPK in each metabolic
T
illustrates the central role of AMPK in MDR, and an intersection analysis between AMPK-
IP
related energy metabolism network and metabolic adaption in resistant cancer cells identifies
the associated group of core genes. Undoubtedly, there are still underlying mechanisms
R
worthy to interpret and many details that need to be worked out, but the core genes identified
SC
here may serve as a guide for future studies, especially for deciphering the metabolic role of
AMPK in MDR. Meanwhile, we hold the opinion that future efforts should be focused on
U
alterations of specific metabolic proteins or pathways in corresponding cancer types, which
N
would provide much more valuable information for personalized medication from the view of
A
omics.
M
Author Contributions
D
preparation with constructive discussion and careful revision. RPC revised the entire
manuscript and edited the language for scientific presentation. YFM and JNL helped to revise
EP
the manuscript and provided valuable feedback to this conception. YTW and TRP conceived
Funding
A
This study was supported by the China Scholarship Council (CSC) and the Fundamental
Conflict of Interest
33
Acknowledgements
We want to thank all the colleagues and friends for their sincere and generous help.
References
T
IP
[1] L.A. Garraway, P.A. Janne, Circumventing cancer drug resistance in the era of
R
personalized medicine, Cancer Discov 2(3) (2012) 214-26.
SC
[2] M. Saraswathy, S. Gong, Different strategies to overcome multidrug resistance in
U
[3] C. Holohan, S. Van Schaeybroeck, D.B. Longley, P.G. Johnston, Cancer drug
N
resistance: an evolving paradigm, Nat Rev Cancer 13(10) (2013) 714-26.
A
[4] K.O. Alfarouk, Tumor metabolism, cancer cell transporters, and
M
[5] R.J. DeBerardinis, N.S. Chandel, Fundamentals of cancer metabolism, Sci Adv 2(5)
D
(2016) e1600200.
TE
[6] R.S. Vidal, J. Quarti, F.D. Rumjanek, V.M. Rumjanek, Metabolic Reprogramming
[7] R.V. Pusapati, A. Daemen, C. Wilson, W. Sandoval, M. Gao, B. Haley, A.R. Baudy,
[8] G.J. Yoshida, Metabolic reprogramming: the emerging concept and associated
34
Vasconcelos, Identification of the metabolic alterations associated with the multidrug
Everts, H. den Dulk, H.S. Overkleeft, B.I. Florea, C. Driessen, Proteasome inhibitor-
adapted myeloma cells are largely independent from proteasome activity and show
T
complex proteomic changes, in particular in redox and energy metabolism, Leukemia
IP
30(11) (2016) 2198-2207.
[11] L. Jerby, E. Ruppin, Predicting drug targets and biomarkers of cancer via genome-
R
scale metabolic modeling, Clin Cancer Res 18(20) (2012) 5572-84.
SC
[12] G. Corona, F. Rizzolio, A. Giordano, G. Toffoli, Pharmaco-metabolomics: an
New Concepts in an Old Field, Antioxid Redox Signal 26(9) (2017) 462-485.
TE
Korbel, M.W. Laschke, P.A. Gimotty, S.E. Philipp, E. Krause, S. Patzold, J. Villanueva,
EP
respiratory chain of slow-cycling JARID1B(high) cells, Cancer Cell 23(6) (2013) 811-
25.
A
[16] V. Koshkin, L.E. Ailles, G. Liu, S.N. Krylov, Metabolic Suppression of a Drug-
Resistant Subpopulation in Cancer Spheroid Cells, J Cell Biochem 117(1) (2016) 59-
65.
35
microenvironment potentiates epithelial to mesenchymal transition and alters cellular
T
[19] D. Tavares-Valente, F. Baltazar, R. Moreira, O. Queiros, Cancer cell bioenergetics
IP
and pH regulation influence breast cancer cell resistance to paclitaxel and doxorubicin,
R
[20] M.A. Pierotti, F. Berrino, M. Gariboldi, C. Melani, A. Mogavero, T. Negri, P.
SC
Pasanisi, S. Pilotti, Targeting metabolism for cancer treatment and prevention:
metformin, an old drug with multi-faceted effects, Oncogene 32(12) (2013) 1475-87.
U
[21] C. Xue, C. Wang, Y. Sun, Q. Meng, Z. Liu, X. Huo, P. Sun, H. Sun, X. Ma, X. Ma,
N
J. Peng, K. Liu, Targeting P-glycoprotein function, p53 and energy metabolism:
A
Combination of metformin and 2-deoxyglucose reverses the multidrug resistance of
M
[22] C. Xue, C. Wang, Q. Liu, Q. Meng, H. Sun, X. Huo, X. Ma, Z. Liu, X. Ma, J. Peng,
D
[23] J.G. Kim, S.J. Lee, Y.S. Chae, B.W. Kang, Y.J. Lee, S.Y. Oh, M.C. Kim, K.H. Kim,
MAPK3/1 expression and prognosis for patients with gastric cancer, Oncology 85(2)
(2013) 78-85.
A
[24] Z. Wang, P. Liu, Q. Chen, S. Deng, X. Liu, H. Situ, S. Zhong, S. Hann, Y. Lin,
Targeting AMPK Signaling Pathway to Overcome Drug Resistance for Cancer Therapy,
[25] D.G. Hardie, AMPK--sensing energy while talking to other signaling pathways,
36
Cell Metab 20(6) (2014) 939-52.
[26] D.G. Hardie, B.E. Schaffer, A. Brunet, AMPK: An Energy-Sensing Pathway with
Multiple Inputs and Outputs, Trends Cell Biol 26(3) (2016) 190-201.
[27] S.P. Mathupala, Y.H. Ko, P.L. Pedersen, Hexokinase-2 bound to mitochondria:
cancer's stygian link to the "Warburg Effect" and a pivotal target for effective therapy,
T
[28] S.P. Mathupala, A. Rempel, P.L. Pedersen, Aberrant glycolytic metabolism of
IP
cancer cells: a remarkable coordination of genetic, transcriptional, post-translational,
and mutational events that lead to a critical role for type II hexokinase, J Bioenerg
R
Biomembr 29(4) (1997) 339-43.
SC
[29] R.G. Coelho, I.C. Calaca, D.M. Celestrini, A.H. Correia-Carneiro, M.M. Costa, P.
[31] W. Yi, P.M. Clark, D.E. Mason, M.C. Keenan, C. Hill, W.A. Goddard, 3rd, E.C.
TE
[32] N.H. Kim, Y.H. Cha, J. Lee, S.H. Lee, J.H. Yang, J.S. Yun, E.S. Cho, X. Zhang,
M. Nam, N. Kim, Y.S. Yuk, S.Y. Cha, Y. Lee, J.K. Ryu, S. Park, J.H. Cheong, S.W.
CC
Kang, S.Y. Kim, G.S. Hwang, J.I. Yook, H.S. Kim, Snail reprograms glucose
[33] B.A. Webb, F. Forouhar, F.E. Szu, J. Seetharaman, L. Tong, D.L. Barber, Structures
37
[34] E. Domenech, C. Maestre, L. Esteban-Martinez, D. Partida, R. Pascual, G.
mitophagy during mitotic arrest, Nat Cell Biol 17(10) (2015) 1304-16.
T
Schulze, Functional metabolic screen identifies 6-phosphofructo-2-kinase/fructose-
IP
2,6-biphosphatase 4 as an important regulator of prostate cancer cell survival, Cancer
R
[36] A. Silvestri, F. Palumbo, I. Rasi, D. Posca, T. Pavlidou, S. Paoluzi, L. Castagnoli,
SC
G. Cesareni, Metformin Induces Apoptosis and Downregulates Pyruvate Kinase M2 in
Breast Cancer Cells Only When Grown in Nutrient-Poor Conditions, PLoS One 10(8)
(2015) e0136250. U
N
[37] D. Dixit, F. Ahmad, R. Ghildiyal, S.D. Joshi, E. Sen, CK2 inhibition induced
A
PDK4-AMPK axis regulates metabolic adaptation and survival responses in glioma,
M
[38] C.A. Wu, Y. Chao, S.G. Shiah, W.W. Lin, Nutrient deprivation induces the Warburg
D
[39] Z. Liu, X. Chen, Y. Wang, H. Peng, Y. Wang, Y. Jing, H. Zhang, PDK4 protein
EP
[40] X. Huang, X. Li, X. Xie, F. Ye, B. Chen, C. Song, H. Tang, X. Xie, High
A
expressions of LDHA and AMPK as prognostic biomarkers for breast cancer, Breast 30
(2016) 39-46.
38
pathway, Biochem J 473(19) (2016) 3013-30.
[42] A.B. Kohan, I. Talukdar, C.M. Walsh, L.M. Salati, A role for AMPK in the
[43] R. Lin, S. Elf, C. Shan, H.B. Kang, Q. Ji, L. Zhou, T. Hitosugi, L. Zhang, S. Zhang,
J.H. Seo, J. Xie, M. Tucker, T.L. Gu, J. Sudderth, L. Jiang, M. Mitsche, R.J.
T
DeBerardinis, S. Wu, Y. Li, H. Mao, P.R. Chen, D. Wang, G.Z. Chen, S.J. Hurwitz, S.
IP
Lonial, M.L. Arellano, H.J. Khoury, F.R. Khuri, B.H. Lee, Q. Lei, D.J. Brat, K. Ye, T.J.
R
oxidative PPP, lipogenesis and tumour growth by inhibiting LKB1-AMPK signalling,
SC
Nat Cell Biol 17(11) (2015) 1484-96.
[44] S.M. Jeon, N.S. Chandel, N. Hay, AMPK regulates NADPH homeostasis to
U
promote tumour cell survival during energy stress, Nature 485(7400) (2012) 661-5.
N
[45] Y.Z. Lee, C.W. Yang, H.Y. Chang, H.Y. Hsu, I.S. Chen, H.S. Chang, C.H. Lee, J.C.
A
Lee, C.R. Kumar, Y.Q. Qiu, Y.S. Chao, S.J. Lee, Discovery of selective inhibitors of
M
Rustin, I. Tomlinson, P.J. Pollard, M.F. Di Renzo, Cells lacking the fumarase tumor
CC
[48] R. Naik, M. Won, H.S. Ban, D. Bhattarai, X. Xu, Y. Eo, Y.S. Hong, S. Singh, Y.
Choi, H.C. Ahn, K. Lee, Synthesis and structure-activity relationship study of chemical
39
[49] H.S. Ban, X. Xu, K. Jang, I. Kim, B.K. Kim, K. Lee, M. Won, A Novel Malate
[50] S. Shah, W.J. Carriveau, J. Li, S.L. Campbell, P.K. Kopinski, H.W. Lim, N. Daurio,
S. Trefely, K.J. Won, D.C. Wallace, C. Koumenis, A. Mancuso, K.E. Wellen, Targeting
T
impinging on an ACLY-AMPK-AR feedback mechanism, Oncotarget 7(28) (2016)
IP
43713-43730.
R
Soga, H. Seimiya, Inhibition of ATP citrate lyase induces an anticancer effect via
SC
reactive oxygen species: AMPK as a predictive biomarker for therapeutic impact, Am
[54] C.J. Halbrook, C.A. Lyssiotis, Employing Metabolism to Improve the Diagnosis
[55] T.T. Hien, H.G. Kim, E.H. Han, K.W. Kang, H.G. Jeong, Molecular mechanism of
suppression of MDR1 by puerarin from Pueraria lobata via NF-kappaB pathway and
CC
activated protein kinase in breast cancer MCF-7/adr cells, Mol Nutr Food Res 54(7)
A
(2010) 918-28.
[56] Z.F. Zhong, W. Tan, W.W. Qiang, V.L. Scofield, K. Tian, C.M. Wang, W.A. Qiang,
human breast cancer cells in an AMPK-dependent manner, Mol Biosyst 12(5) (2016)
40
1626-37.
T
kinase by retinoic acid sensitizes hepatocellular carcinoma cells to apoptosis induced
IP
by sorafenib, Cancer Sci 106(5) (2015) 567-75.
[59] C.H. Chang, C.Y. Lee, C.C. Lu, F.J. Tsai, Y.M. Hsu, J.W. Tsao, Y.N. Juan, H.Y.
R
Chiu, J.S. Yang, C.C. Wang, Resveratrol-induced autophagy and apoptosis in cisplatin-
SC
resistant human oral cancer CAR cells: A key role of AMPK and Akt/mTOR signaling,
[61] Y. Wu, Y. Qi, H. Liu, X. Wang, H. Zhu, Z. Wang, AMPK activator AICAR
D
promotes 5-FU-induced apoptosis in gastric cancer cells, Mol Cell Biochem 411(1-2)
TE
(2016) 299-305.
[62] S. Ling, Y. Tian, H. Zhang, K. Jia, T. Feng, D. Sun, Z. Gao, F. Xu, Z. Hou, Y. Li,
EP
[63] C. Qu, W. Zhang, G. Zheng, Z. Zhang, J. Yin, Z. He, Metformin reverses multidrug
protein kinase (AMPK) in human breast cancer cells, Mol Cell Biochem 386(1-2)
(2014) 63-71.
[64] H.G. Kim, T.T. Hien, E.H. Han, Y.P. Hwang, J.H. Choi, K.W. Kang, K.I. Kwon,
B.H. Kim, S.K. Kim, G.Y. Song, T.C. Jeong, H.G. Jeong, Metformin inhibits P-
41
glycoprotein expression via the NF-kappaB pathway and CRE transcriptional activity
T
13(7) (2014) 1132-44.
IP
[66] M.B. Aldonza, J.Y. Hong, S.Y. Bae, J. Song, W.K. Kim, J. Oh, Y. Shin, S.H. Lee,
R
Associated Multidrug Resistance by 21alpha-Methylmelianodiol in Lung Cancer Cells,
SC
PLoS One 10(6) (2015) e0127841.
[67] C. Ji, B. Yang, Y.L. Yang, S.H. He, D.S. Miao, L. He, Z.G. Bi, Exogenous cell-
U
permeable C6 ceramide sensitizes multiple cancer cell lines to Doxorubicin-induced
N
apoptosis by promoting AMPK activation and mTORC1 inhibition, Oncogene 29(50)
A
(2010) 6557-68.
M
inhibitor of hippo, TGF-beta, and Wnt signaling pathways, Mol Cancer Ther 13(6)
(2014) 1457-67.
CC
[70] W.D. Wu, Z.M. Hu, M.J. Shang, D.J. Zhao, C.W. Zhang, D.F. Hong, D.S. Huang,
[71] T.P. Tran, H.G. Kim, J.H. Choi, M.K. Na, H.G. Jeong, Reversal of P-glycoprotein-
42
Phytomedicine 20(7) (2013) 622-31.
[72] K.H. Eum, S.K. Ahn, H. Kang, M. Lee, Differential inhibitory effects of two Raf-
[73] J.H. Ahn, M. Lee, Suppression of autophagy sensitizes multidrug resistant cells
towards Src tyrosine kinase specific inhibitor PP2, Cancer Lett 310(2) (2011) 188-97.
T
[74] Q. Zhang, Y. Feng, D. Kennedy, Multidrug-resistant cancer cells and cancer stem
IP
cells hijack cellular systems to circumvent systemic therapies, can natural products
R
[75] D. Uribe, A. Torres, J.D. Rocha, I. Niechi, C. Oyarzun, L. Sobrevia, R. San Martin,
SC
C. Quezada, Multidrug resistance in glioblastoma stem-like cells: Role of the hypoxic
[77] A. Schulze, A.L. Harris, How cancer metabolism is tuned for proliferation and
[78] L.K. Boroughs, R.J. DeBerardinis, Metabolic pathways promoting cancer cell
TE
M. Simonovic, A. Roth, A. Santos, K.P. Tsafou, M. Kuhn, P. Bork, L.J. Jensen, C. von
Mering, STRING v10: protein-protein interaction networks, integrated over the tree of
CC
Santos, N.T. Doncheva, A. Roth, P. Bork, L.J. Jensen, C. von Mering, The STRING
[81] A.J. Enright, S. Van Dongen, C.A. Ouzounis, An efficient algorithm for large-scale
43
detection of protein families, Nucleic Acids Res 30(7) (2002) 1575-84.
[82] P. Shannon, A. Markiel, O. Ozier, N.S. Baliga, J.T. Wang, D. Ramage, N. Amin,
[83] M. Talekar, Q. Ouyang, M.S. Goldberg, M.M. Amiji, Cosilencing of PKM-2 and
T
Model of Ovarian Cancer, Mol Cancer Ther 14(7) (2015) 1521-31.
IP
[84] J.R. Mayers, M.G. Vander Heiden, Famine versus feast: understanding the
R
[85] E. Cerami, J. Gao, U. Dogrusoz, B.E. Gross, S.O. Sumer, B.A. Aksoy, A. Jacobsen,
SC
C.J. Byrne, M.L. Heuer, E. Larsson, Y. Antipin, B. Reva, A.P. Goldberg, C. Sander, N.
Schultz, The cBio cancer genomics portal: an open platform for exploring
U
multidimensional cancer genomics data, Cancer Discov 2(5) (2012) 401-4.
N
[86] J. Gao, B.A. Aksoy, U. Dogrusoz, G. Dresdner, B. Gross, S.O. Sumer, Y. Sun, A.
A
Jacobsen, R. Sinha, E. Larsson, E. Cerami, C. Sander, N. Schultz, Integrative analysis
M
of complex cancer genomics and clinical profiles using the cBioPortal, Sci Signal
[87] B.E. Johnson, T. Mazor, C. Hong, M. Barnes, K. Aihara, C.Y. McLean, S.D. Fouse,
TE
S. Yamamoto, H. Ueda, K. Tatsuno, S. Asthana, L.E. Jalbert, S.J. Nelson, A.W. Bollen,
W.C. Gustafson, E. Charron, W.A. Weiss, I.V. Smirnov, J.S. Song, A.B. Olshen, S. Cha,
EP
Y. Zhao, R.A. Moore, A.J. Mungall, S.J.M. Jones, M. Hirst, M.A. Marra, N. Saito, H.
Aburatani, A. Mukasa, M.S. Berger, S.M. Chang, B.S. Taylor, J.F. Costello, Mutational
CC
analysis reveals the origin and therapy-driven evolution of recurrent glioma, Science
44
Cheng, H. Xue, Y. Wang, D. Lin, A.J. Mungall, R. Moore, Y. Zhao, J. Lorette, L.
Nguyen, D. Huntsman, C.J. Eaves, C. Hansen, M.A. Marra, C. Caldas, S.P. Shah, S.
T
Kelley, J.A. Sosman, D.B. Johnson, A. Ribas, R.S. Lo, Genomic and Transcriptomic
IP
Features of Response to Anti-PD-1 Therapy in Metastatic Melanoma, Cell 165(1)
(2016) 35-44.
R
[90] A.L. Cohen, S.L. Holmen, H. Colman, IDH1 and IDH2 mutations in gliomas, Curr
SC
Neurol Neurosci Rep 13(5) (2013) 345.
[91] L.E. Jalbert, A. Elkhaled, J.J. Phillips, E. Neill, A. Williams, J.C. Crane, M.P.
U
Olson, A.M. Molinaro, M.S. Berger, J. Kurhanewicz, S.M. Ronen, S.M. Chang, S.J.
N
Nelson, Metabolic Profiling of IDH Mutation and Malignant Progression in Infiltrating
A
Glioma, Sci Rep 7 (2017) 44792.
M
[92] H. Yan, D.W. Parsons, G. Jin, R. McLendon, B.A. Rasheed, W. Yuan, I. Kos, I.
Herndon, K.W. Kinzler, V.E. Velculescu, B. Vogelstein, D.D. Bigner, IDH1 and IDH2
TE
[93] M.S. Waitkus, B.H. Diplas, H. Yan, Biological Role and Therapeutic Potential of
EP
[94] B.M. Lindqvist, S. Wingren, P.B. Motlagh, T.K. Nilsson, Whole genome DNA
CC
62.
A
[95] X. Ge, Z. Cao, Y. Gu, F. Wang, J. Li, M. Han, W. Xia, Z. Yu, P. Lyu, PFKFB3
(2016) 119-25.
45
[96] M.J. Kim, D.H. Kim, W.H. Jung, J.S. Koo, Expression of metabolism-related
proteins in triple-negative breast cancer, Int J Clin Exp Pathol 7(1) (2014) 301-12.
the nucleus in MCF7 breast cancer cells, J Proteome Res 11(12) (2012) 6080-101.
[98] T. Contractor, C.R. Harris, p53 negatively regulates transcription of the pyruvate
T
dehydrogenase kinase Pdk2, Cancer Res 72(2) (2012) 560-7.
IP
[99] E. Papaevangelou, G.S. Almeida, C. Box, N.M. deSouza, Y.L. Chung, The effect
R
carcinoma model, Int J Cancer 143(4) (2018) 992-1002.
SC
[100] S. Al-Bahlani, H. Al-Lawati, M. Al-Adawi, N. Al-Abri, B. Al-Dhahli, K. Al-
Adawi, Fatty acid synthase regulates the chemosensitivity of breast cancer cells to
U
cisplatin-induced apoptosis, Apoptosis 22(6) (2017) 865-876.
N
[101] J.A. Menendez, R. Lupu, Fatty acid synthase (FASN) as a therapeutic target in
A
breast cancer, Expert Opin Ther Targets 21(11) (2017) 1001-1016.
M
[102] H. Liu, X. Huang, T. Ye, MiR-22 down-regulates the proto-oncogene ATP citrate
lyase to inhibit the growth and metastasis of breast cancer, Am J Transl Res 10(3)
D
(2018) 659-669.
TE
[103] K.S. Lucenay, I. Doostan, C. Karakas, T. Bui, Z. Ding, G.B. Mills, K.K. Hunt, K.
Enable Malignant Growth of Breast Cancer Cells, Cancer Res 76(8) (2016) 2406-18.
46
characterization of altered membrane lipid metabolism in breast cancer progression,
Gao, W. Ritchie, Y. Feng, C.G. Bailey, N. Deng, K. Harvey, J.M. Beith, C.I. Selinger,
S.A. O'Toole, J.E. Rasko, J. Holst, ASCT2/SLC1A5 controls glutamine uptake and
T
3201-8.
IP
[107] Y. Wang, S. Fan, J. Lu, Z. Zhang, D. Wu, Z. Wu, Y. Zheng, GLUL Promotes Cell
R
[108] P. Malvi, B. Chaube, V. Pandey, M.V. Vijayakumar, P.R. Boreddy, N. Mohammad,
SC
S.V. Singh, M.K. Bhat, Obesity induced rapid melanoma progression is reversed by
orlistat treatment and dietary intervention: role of adipokines, Mol Oncol 9(3) (2015)
689-703. U
N
[109] V. Pandey, M.V. Vijayakumar, A.K. Ajay, P. Malvi, M.K. Bhat, Diet-induced
A
obesity increases melanoma progression: involvement of Cav-1 and FASN, Int J Cancer
M
[110] J. Wu, J. Du, X. Fu, B. Liu, H. Cao, T. Li, T. Su, J. Xu, A.K. Tse, Z.L. Yu, Iciartin,
D
[111] F. Vazquez, J.H. Lim, H. Chim, K. Bhalla, G. Girnun, K. Pierce, C.B. Clish, S.R.
EP
[113] Y. Zhao, E.B. Butler, M. Tan, Targeting cellular metabolism to improve cancer
47
[114] W. Tan, Z. Zhong, S. Wang, H. Liu, H. Yu, R. Tan, X. Hu, T. Pan, Y. Wang, The
T
R IP
SC
U
N
A
M
D
TE
EP
CC
A
48
Figure Captions
The specific interactions of AMPK (represented as a grouped circle comprised of seven key
proteins) in each identified metabolic process, including glycolysis (A), OXPHOS related
T
process (B), fatty acid synthesis related process (C) and glutamine related process (D). Nodes
IP
represent proteins and edges (the lines between two nodes) represent protein-proteins
R
associations. The size of the nodes indicates the level of functional associations between
SC
proteins according to number of database annotations. Node color was derived from co-
expression data. Edge color represents the variance of interactions parsed across primary
PRKAB2, PRKAG1, PRKAG2, and PRKAG3) were grouped by heat maps with the
computed text-mining scores, elucidating these closest interactions between AMPK and
D
glycolysis related proteins (A), OXPHOS related proteins (B), fatty acid synthesis related
TE
Figure 3 The central role of AMPK and the complex relationships between these four
metabolic processes
CC
These directly linked proteins comprise a heterogeneous group with 22 members indicate a
central role of AMPK in energy metabolism network (A), and a Venn diagram visualization
A
of the complex relationships between these four groups, with the proportion of shared
49
The key proteins (165 proteins) derived from AMPK-based network through data mining and
the altered proteins (137 proteins) in metabolic adaptation are compared. 28 proteins were
found at this intersection, accounting for 10.2% of the total. These identified core proteins
Figure 5 Gene networks analysis of 28 genes responsible for encoding the core proteins
T
(A) Greater than 30% of altered genes in these core 28 genes is criteria, hence 57
IP
studies were found according to the cancer genome atlas (TCGA) provisional
R
data. With quite remarkable significance across 166 various studies, a high
SC
alteration frequency (>80.0%) ranked three cancer types: glioma (100%), breast
cancer (96.6%) and melanoma (80.6%). For each caner type, genetic alterations
U
for each gene involved in AMPK-mediated metabolic pathways of MDR, which
N
were found in at least three studies, are visualized according to their percent
A
alteration frequency, including glioma (B), breast cancer (C) and melanoma (D).
M
D
TE
EP
CC
A
50
A
Fig 1B
Fig 1 A
CC
EP
TE
D
M
A
N
U
SC
R
51
IP
T
A Fig 1 C
Fig 1 D
CC
EP
TE
D
M
A
N
U
SC
R
52
IP
T
A
Fig 2B
Fig 2 A
CC
EP
TE
D
M
A
N
U
SC
R
53
IP
T
A Fig 2 C
Fig 1 D
CC
EP
TE
D
M
A
N
U
SC
R
54
IP
T
A
Fig 3B
Fig 3 A
CC
EP
TE
D
M
A
N
U
SC
R
55
IP
T
A
CC
EP
TE
D
M
A
N
U
SC
R
56
IP
T
A Fig 4
CC
EP
TE
D
M
A
N
U
SC
R
57
IP
T
A
Fig 5 B
Fig 5 A
CC
EP
TE
D
M
A
N
U
SC
R
58
IP
T
A
Fig D
Fig 5C
CC
EP
TE
D
M
A
N
U
SC
R
59
IP
T