Talanta 150 (2016) 8896

Contents lists available at ScienceDirect

Talanta
journal homepage: www.elsevier.com/locate/talanta

High resolution mass spectrometry coupled with multivariate data

analysis revealing plasma lipidomic alteration in ovarian cancer in
Asian women
Yangyang Zhang a, Yingying Liu b, Lin Li a,c, Jinchao Wei a, Shaoxiang Xiong a,
Zhenwen Zhao a,c,n
a
Beijing National Laboratory for Molecular Sciences, Key Laboratory of Analytical Chemistry for Living Biosystems, Institute of Chemistry Chinese Academy of
Sciences, Beijing Mass Spectrum Center, Beijing 100190, China
b
Department of Obstetrics and Gynecology, Peking University Shougang Hospital, Beijing 100144, China
c
Graduate School, University of Chinese Academy of Sciences, Beijing 100049, China

art ic l e i nf o a b s t r a c t

Article history: Ovarian cancer (OC) is the most common cause of death from gynecologic malignancies in women. The
Received 23 October 2015 identication of reliable diagnostic biomarkers for the early detection of this deadly disease is critical for
Received in revised form reducing the mortality rate of OC. Plasma lysophosphatidic acid (LPA) levels were increased from OC
25 November 2015
patients vs. healthy controls. Therefore, lipidomics may represent an excellent developing prospect for
Accepted 10 December 2015
Available online 11 December 2015
the discovery of diagnostic biomarkers of OC. In this study, a nontargeted lipidomics approach based on
ultra performance liquid chromatography-electrospray ionization-QTOF-mass spectrometry (UPLC-ESI-
Keywords: QTOF-MS) combined with multivariate data analysis, including principal component analysis (PCA) and
Mass spectrometry (orthogonal) partial least squared discriminant analysis [(O)PLS-DA] was applied for the investigation of
Lipidomics
potential diagnostic biomarkers in plasma of OC patients. Patients with OC could be distinguished from
Ovarian cancer
healthy individuals and patients with benign gynecological tumor disease by this method, which shows a
Multivariate data analysis
Glycerophospholipid metabolism signicant lipid perturbation in this disease. With the assistance of high resolution and high accuracy of
MS and MS/MS data, the potential markers including lysophosphatidylcholines (LPCs), phosphati-
dylcholines (PCs) and triacylglycerols (TGs) with specic fatty acid chains, were identied. Interestingly,
LPCs were up-regulated and PCs and TGs were down-regulated, compared OC group with benign tumor
and normal control groups, and the glycerophospholipid metabolism emerged as a key pathway, in
particular, the phospholipase A2 (PLA2) enzyme activity, that was disregulated in the disease. This study
may provide new insight into underlying mechanisms for OC and proves that MS-based lipidomics is a
powerful method in discovering new potential clinical biomarkers for diseases.
& 2015 Elsevier B.V. All rights reserved.

1. Introduction
Ovarian cancer (OC) is the most common cause of death from
gynecologic malignancies in women of all ages in the Western
world and caused an estimated 14,270 deaths in 2014 in the
United States [1]. Early stage (I/II) of OC has an excellent prognosis
with a survival rate of over 90%, but approximate 80% of all re-
ported cases are caught in the advanced stages (III/IV) with the
5-year survival rate at only 11% [2,3]. Given the success of treat-
ment for early stage disease, a reliable diagnostic marker for OC at
an early stage is intuitively attractive. The highly glycosylated CA-
n
Corresponding author at: Beijing National Laboratory for Molecular Sciences,
Key Laboratory of Analytical Chemistry for Living Biosystems, Institute of Chemistry
Chinese Academy of Sciences, Beijing Mass Spectrum Center, Beijing 100190, China.
E-mail address: zhenwenzhao@iccas.ac.cn (Z. Zhao).
125 protein is a FDA approved blood test for the diagnosis of OC;
however, this marker lacks the sensitivity and specicity to be
used either as an early warning marker or for population screening
[47]. HE4, a new biomarker for OC patients, shows similar sen-
sitivity and specicity with CA125 in the diagnosis of ovarian
cancer patients, having a slightly better efcacy in premenopausal
patients [810]. Thus, the identication of reliable diagnostic
biomarkers for the early detection of this deadly disease is critical
for reducing the mortality rate of OC.
Lipidomics is a growing research eld that studies cellular li-
pidomes on a large scale and at the intact-molecule level. Recent
accumulating evidence suggests that cancer and metabolic dis-
eases are connected with the dysregulation of lipid metabolism
[11]. It was reported that lysophosphatidic acid (LPA) and lyso-
phosphotidylinositol (LPI) are signicantly increased in OC ascites
compared with nonmalignant disease ascites (e.g., liver failure)

http://dx.doi.org/10.1016/j.talanta.2015.12.021
0039-9140/& 2015 Elsevier B.V. All rights reserved.
 Y. Zhang et al. / Talanta 150 (2016) 8896
[12]. In addition, ample evidences support plasma LPA levels were
increased from OC patients vs. healthy controls [13,14]. Therefore,
lipidomics represents an excellent developing prospect for the
discovery of diagnostic biomarkers of OC [1518].
Multiple recent advances in mass spectrometric approaches
have greatly extended the analytical capabilities to facilitate the
accurate analysis of global lipids to provide new insights into li-
pidomic pathways and functions [1922]. A lipidomic LC/MS
platform based on chromatographic retention, high mass resolu-
tion and accuracy, MS/MS analysis, and quantitation software en-
ables analysis of complex samples by LTQ-FT MS for lipid droplets,
nding enhanced sensitivity for detection of PE and DG species
[23]; Plasma metabolomic investigation of hepatocellular carci-
noma patients was reported by ultra performance liquid chroma-
tography-electrospray ionization-quadrupole time-of-ight mass
spectrometry, downregulated molecules of interest included 11
LPCs in the plasma of HCC patients provide new insights into the
pathobiology of the disease [16]; a uniform DHB layer-assisted
MALDI-FTICR MS-based approach was developed for the rapid
analysis of lipid extracts from cell pellets, which showed higher
throughput and is more convenient, presenting an alternative
approach for lipidomics study [21]. Thus, MS-based lipidomic
analyses have signicantly expanded our knowledge related to
human physiology and pathology [24].
In this study, a nontargeted lipidomics approach based on ultra
performance liquid chromatography-electrospray ionization-
quadrupole time-of-ight-mass spectrometry (UPLC-ESI-QTOF-
MS) in conjunction with multivariate data analysis, including
principal component analysis (PCA) and (orthogonal) partial least
squared discriminant analysis [(O)PLS-DA] was applied for the
rapid investigation of potential diagnostic biomarkers in plasma of
OC patients. The high resolution mass spectrometry and its cor-
responding MS/MS data were combined to identify the structure of
these potential biomarkers. Additionally, these data were as well
uploaded to MetaboAnalyst website (www.metaboanalyst.ca) for
pathways analysis using MetPA software to identify the dis-
regulated pathways.
2. Experimental
2.1. Materials
Lipid standards, 1-lauroyl-2-hydroxy-sn-glycero-3-phos-
phocholine (12:0 LPC) and 1-myristoyl-2-hydroxy-sn-glycero-3-
phosphate (14:0 LPA) were purchased from Avanti Polar Lipids
(Birmingham, AL). HPLC-grade methanol (MeOH), acetonitrile
(CH3CN), isopropanol (IPA), chloroform(CHCl3), formic acid as well
as ammonium formate were purchased from Sigma or Fisher Sci-
entic. Ultrapure water from Milli-Q purication system (Millipore
Corporation, USA) was used. All of the above materials were used
as-received condition without further purication.
2.2. Sample collection and lipids extraction
2.2.1. Participants
Plasma samples of Asian women were a gift from Dr. Yingying
Liu, including 27 patients with OC (C group, age range: 3774,
55.2 710.7 years old), 27 patients with benign gynecological tu-
mor (Artful bursa benign tumor, B group, age range: 876,
47.2 717.6 years old), and 11 unaffected controls (N group, age
range: 3071, 45.9 713.1 years old) enrolled at the Shougang
Hospital of Peking University. Blood samples were collected in
EDTA-containing tubes and centrifuged at 1750
g for 15 min at
room temperature. Plasma samples were aliquoted into siliconized
eppendorf tubes (SafeSeal microcentrifuge tubes; PGC Scientics,
89
Frederick, MD) and frozen at 80 C until use. The project was
approved by the Institutional Review Board, and written informed
consent forms were signed by participants.
2.2.2. Lipids extraction
Lipids were extracted by methanol method [25]. In brief, 20 l
of plasma was transferred into 1000 L of methanol containing
12:0 LPC (1 nmol, used for quality control in positive-ion detection
mode) and 14:0 LPA (20 pmol, used for quality control in negative-
ion detection mode). After vortexing and centrifugation (10,000 g,
5 min, room temperature), 2 L of the supernatant or 2 L of 10X
dilution were loaded into mass spectrometer for negative-ion or
positive-ion lipids analysis, respectively.
2.3. UPLC-ESI-QTOF mass spectrometry
Lipids prole analysis was performed using UPLC-ESI-QTOF-MS
(Xevo G2 Q-TOF MS, Waters). Samples were loaded through a LC
system (I-class Acquity ultra performance liquid chromatography,
Waters) with an auto sampler. The mobile phase A was iso-
propanol/acetonitrile/formic acid (90:10:0.1, v/v/v) containing
10 mM ammonium formate; The mobile phase B was acetonitrile/
water/formic acid (70:30:0.1, v/v/v) containing 10 mM ammonium
formate. A CSH C18 column (1.7 m, 2.1 mm ID
100 mm, Waters)
was used for separation of lipids. The column was maintained at
55 C. The UPLC separations were 20 min/sample using the fol-
lowing scheme: (1) 0 min, 70% B; (2) 2 min, 57% B; (3) 2.1 min, 50%
B; (4) 12 min, 46% B; (5) 12.1 min, 30% B; (6) 18 min, 1% B;
(7) 18.1 min, 70% B; (8) 20 min, 70% B. All the changes are linear,
and the ow rate was set to 400 L/min.
The MS system was operated using both the positive-ion
(ESI ) and negative-ion (ESI-) modes. The m/z range was set at
1001500. The ionization source temperature was set at 110 C.
Nitrogen was used as the dry gas and collision gas. The cone gas
ow was set at 15 L/h, and the desolvation gas ow was set at
600 L/h, and the desolvation gas temperature was set at 320 C. In
positive ion mode, the capillary voltage was 3.5 kV, the sampling
cone voltage was 35 V, and the extraction cone voltage was 3.0 V.
In negative ion mode, the capillary voltage was 2.8 kV, the sam-
pling cone voltage was 35 V, and the extraction cone voltage was
3.0 V, Multiplexed data acquisition (MSE), a kind of data in-
dependent acquisition modes was performed for simultaneously
acquiring MS and MS/MS data with high resolution and high ac-
curacy. MS/MS data were obtained for all the ions observed in the
preceding MS scan. The lock spray was used to ensure reprodu-
cibility and accuracy. Leucine enkephalin (200 pg/ml) was used as
lock mass in ESI- (m/z 554.2615) and ESI (m/z 556.2615). The
lock spray frequency was set to 5 s, and the lock mass data were
averaged over a total of 10 scans.
2.4. Data preprocessing
Data obtained by QTOF were processed by Markerlynx XS and
EZ info software (Waters). Firstly, the intensity of each ion was
normalized to generate a data matrix that consisted of the m/z
value, the retention time, and the normalized peak area. Next,
unsupervised segregation was assessed with principal compo-
nents analysis (PCA) using pareto-scaled data. PCA data were vi-
sualized by plotting the PCA scores; each point represents one
retention time/mass-to-charge (m/z) pair. Thus, the loadings plot
gives an indication of the lipids that most strongly inuence the
patterns in the score plot. To maximize class discrimination and
biomarkers identication, the data were further analyzed using
the (O)PLS-DA method, whereas S-plots or loadings plot were
calculated to visualize the relationship between covariance and
correlation within the (O)PLS-DA data. Discriminating variables
90 Y. Zhang et al. / Talanta 150 (2016) 8896

were selected according to variable importance in projection va- A

6000000 glycerophospholipids
lues (VIP) 45.0. Furthermore, an independent t test (p o0.05) and sphingolipids
(Microsoft Ofce Excel 2007) was used to determine if different
biomarker candidates obtained from OPLS-DA modeling were 5000000

statistically signicant between groups.

lysophospholipids
4000000
2.5. Biomarker identication

Intensity(cps)
3000000
The markers were identied by searching local database es-
tablished by Waters or free online databases, including Lipidmaps 12:0 LPC
(http://www.lipidmaps.org/) and the Human Metabolome Data- 2000000
triacylglycerols
base (HMDB) (http://www.hmdb.ca), to precisely match the ac-
curate molecular ion obtained by MS (mass error o5 ppm), and 1000000
the fragment ions obtained by MS/MS were further utilized for
conrmation of the marker.
0
0 5 10 15 20
2.6. Construction of lipidomic pathways Time(min)

The construction, interaction, and pathway analysis of potential B

glycerophospholipids
biomarkers of OC patients was performed with Metaboanalyst 2.0. lysophospholipids
and sphingolipids
140000
This program, based on several databases, including KEGG (http://
www.genome.jp/kegg/) and the Human Metabolome Database 120000 14:0 LPA
(http://www.hmdb.ca/), helps identify pathways that are most
signicantly altered. Possible biological roles were also evaluated 100000
Intensity(cps)
using enrichment analysis of Metaboanalyst 2.0.
80000

3. Results and discussions 60000

3.1. Prole analysis of lipids in plasma by UPLC-ESI-QTOF-MS
Using the UPLC-ESI-QTOF-MS described above, we performed
prole analysis of lipids in plasma in both positive and negative
ion detection modes. The UPLC conditions were optimized. Several
columns, like BEH C18 (1.7 m, 2.1 mm
100 mm, Cat. No.:
186002352), HILIC (1.7 m, 2.1 mm
100 mm, Cat. No.:
186003461), HSS T3 (1.8 m, 2.1 mm
100 mm, Cat. No.:
186003539) and CSH C18 (1.7 m, 2.1 mm
100 mm Cat. No.:
186005297), and different mobile phases were tested. Basically,
the CSH C18 column provided an acceptable separation capability
for lipids analysis [20]. The representative total ion chromato-
grams (TICs) of lipid extracts from plasma sample were shown in
Fig. 1. The lysophospholipids were eluted before 2 min, glycer-
ophospholipids and sphingolipids were eluted between 5 and
13 min, and triacylglycerols (TGs) were eluted after 15 min. The
relative standard derivation (RSD) of peak area of 12:0 LPC in
positive-ion detection mode or 14:0 LPA in negative-ion detection
mode was 4.6% or 7.8%, respectively, showing the stability and
reproducibility of our MS-based lipidomics approach.
3.2. (O)PLS-DA for discrimination of ovarian cancer patients (C
group), benign ovarian tumor patients (B group) and normal in-
dividuals (N group)
Markerlynx XS was used to carry out peak discrimination, l-
tering, and retention time alignment, yielding 44000 peaks be-
tween retention time of 0.220 min. To evaluate the capability of
the UPLC-ESI-QTOF-MS-based lipidomic approach to differentiate
ovarian cancer patients (C group), benign ovarian tumor patients
(B group) and normal individuals (N group), principal component
analysis (PCA) was performed. In the PCA score plots, differences
among C group, B group and N group were observed. However, the
C, B, and N groups were not well distinguished in the PCA score
plot (data not shown). This is likely because human plasma sam-
ples are extremely complex, and the PCA data analysis technique
separates samples based on unsupervised, random, OC-irrelevant
40000
20000
0
0 5 10 15 20
Time(min)
Fig. 1. The representative total ion chromatograms (TICs) of lipids extracts from
plasma obtained by UPLC-ESI( )-QTOF-MS (A) and UPLC-ESI(  )-QTOF-MS (B).
variation of lipids.
To further investigate the global lipids alterations among C, B,
and N groups, all observations acquired were integrated and co-
analyzed using supervised OPLS-DA or PLS-DA. OPLS-DA can be
used for differentiate 2 groups; while PLS-DA can be used for
differentiate more than 2 groups. As presented in Fig. 2A and C,
and Fig. 3A and C, clear separation by OPLS-DA, based on the data
from UPLC-ESI( ) or ESI(  )-QTOF-MS, was observed between C
and B groups, and between C and N groups, suggesting that lipid
metabolic perturbations were evident in the samples dependent
on pathological condition. Further, based on the data from UPLC-
ESI-QTOF-MS, PLS-DA was performed to differentiate C, B and N
groups. It was showed that the data from UPLC-ESI( )-QTOF-MS
did not completely separate these 3 groups, and the PLS-DA score
plot was shown in Fig. 4A; however, the data from UPLC-ESI
( )-QTOF-MS did completely separate these 3 groups, and the
PLS-DA score plot was shown in Fig. 4C.
The VIP analysis in Markerlynx XS was further carried out to
select variables as primary markers for distinguishing patients
from controls. The VIP value reects the inuence of every variable
on the classication. In general, variables with a VIP value 41.0
had an above average inuence. In our work, the variables with a
VIP value 45.0 were used, and the corresponding S-plot, Fig. 2B
and Fig. 2D, and Fig. 3B and Fig. 3D, and loadings plot, Fig. 4B and
Fig. 4D, were shown respectively. In S-plot or loadings plot, each
point represents a variable with specic retention time and m/z
value. The variables with VIP value 45.0 were highlighted with
 Y. Zhang et al. / Talanta 150 (2016) 8896 91

Fig. 2. OPLS-DA score plots (A and C) and S-plots (B and D) based on the data from UPLC-ESI( )-QTOF-MS (A and B) or UPLC-ESI(  )-QTOF-MS (C and D) for distinguishing C
group and N group, and for selecting the potential markers. The variables with VIP value 45.0 were highlighted with red boxes. (For interpretation of the references to color
in this gure legend, the reader is referred to the web version of this article.)

red boxes, and the variables with specic retention time and m/z
values were hence selected. In summary, a total of 18, 22 and 25
variables were selected for distinguishing C and N groups, C and B
group, and C, B and N groups respectively in positive ion detect
mode; and similarly, a total of 18, 19, and 16 variables were se-
lected for distinguishing C and N groups, C and B group, and C, B
and N groups respectively in negative ion detect mode. These
variables have the potential to distinguish patients from controls,
and therefore merit further studies to identify the structures of
these potential markers, which would provide the basis for future
study on function of these compounds.
3.3. Identication of potential markers for discrimination of C, B and
N groups.
To identify the structures of these potential markers, their m/z
information obtained by positive and negative-ion detection mode
was used to search the matched lipid from database established by
Waters or from LIPID MAPS or HMDB. The error was set at 5 ppm.
The structure was further conrmed by MS/MS data.
In positive ion detect mode, a variable with RT at 5.72 min and
m/z at 784.5838 was selected as primary markers for distin-
guishing C and N group. The mass spectrum with RT at 5.72 in
positive-ion detection mode showed 2 main signals with m/z at
784.5856 and 806.5667, corresponding to the [MH] and
[MNa] . The database (LIPID MAPS) search results for the given
m/z at 784.5838 were 36:3 PC or 49:3 PE with a mass error
1.65 ppm. The MS/MS data (Fig. 5A) with RT at 5.72 min showed
that an ion with m/z at 184.0740 (phosphorylcholine) was the
main fragment ion. Based on our previous studies, it can be in-
ferred that this ion (m/z: 784.5838) was likely a phosphati-
dylcholine (PC). In negative ion detect mode, in the MS data with
RT at 5.72 min, a quasimolecular ion with m/z at 828.5766 was
shown. In the mobile phase, an ammonium formate was added as
additive to enhance the ionization efciency. Therefore, the ion
with m/z at 828.5766 was likely the quasimolecular ion
[MHCOO]  . The database (HMDB) search results for the given
m/z at 828.5766 were also showed that this compound was likely
36:3 PC with a mass error 1.45 ppm. The MS/MS data (Fig. 5B) with
RT at 5.72 min showed that two ions with m/z at 255.2328
(C15H31COO  ) and at 305.2486 (C19H34COO  ), corresponding to
two fatty acid chain linked to glycerol backbone in the structure of
PC, were the main fragment ions. All together, we inferred that the
variable with RT at 5.72 min was 16:0/20:3 PC.
In positive ion detect mode, 2 variables with RT at around
15.04 min and m/z at 872.7671 and 898.7820 were selected as
primary markers for distinguishing C and B group. The mass
spectrum with RT at 15.04 in positive-ion detection mode (Fig. 6A)
92 Y. Zhang et al. / Talanta 150 (2016) 8896

Fig. 3. OPLS-DA score plots (A and C) and S-plots (B and D) based on the data from UPLC-ESI( )-QTOF-MS (A and B) or UPLC-ESI(  )-QTOF-MS (C and D) for distinguishing C
group and B group, and for selecting the potential markers. The variables with VIP value 4 5.0 were highlighted with red boxes. (For interpretation of the references to color
in this gure legend, the reader is referred to the web version of this article.)

showed a serial of signals with m/z at 872.7671, 877.7273,
893.6942 and 898.7820, 903.7374. It could be inferred that
2 compounds were simultaneously eluted from UPLC column, and
these signals were corresponding to the [M1 NH4] , [M1 Na] ,
[M1 K] and [M2 NH4] , [M2 Na] (M1 and M2 were 2 dif-
ferent molecules). The database (LIPID MAPS) search results for
the given m/z at 872.7671 and 898.7820 were 52:4 TG (M1) and
54:5 TG (M2) with a mass error 3.44 ppm and 4.56 ppm, re-
spectively. The MS/MS data with RT at 15.04 min was shown in
Fig. 6B. In Fig. 6B, due to MSE mode was adopted for acquisition of
MS/MS data, the fragment ions were mixed and from the ions
eluted together from UPLC column. Carefully investigating these
data, we could infer that the fragment ions with m/z at 575.5032
([M1 NH4-17-C18H32O2] ), 599.5023 ([M1 NH4-17-C16H32O2]
or [M2 NH4-17-C18H34O2] ), 601.5072 ([M2 NH4-17
-C18H32O2] ) were formed by neutral loss (NL) of NH3 and fatty
acids from amounium adduct triacylglycerol. All together, we in-
ferred that the 2 variables with RT at 15.04 min were 18:2/18:2/
16:0 TG and 18:2/18:1/18:1 TG, respectively.
Similarly, we identied some other potential markers. In total,
15 lipids were identied as possible biomarkers for distinguishing
C group and N group (Table 1), 18 lipids were identied as possible
biomarkers for distinguishing C group and B group (Table 2), and
11 lipids were identied for distinguishing C group, N group and B
group (Table 3). The more detailed information, including fold
change (FC, the ratio of lipid level in C group versus B or N group),
p value, etc. was listed in Tables 13. Basically, the mainly identi-
ed potential markers were lysophosphatidylcholines (LPCs),
phosphatidylcholines (PCs) and triacylglycerols (TGs). The vari-
ables simultaneously identied by both positive and negative ion
detect modes showed very consistent trend. For example, the fold
change (FC) of 18:1 LPC were 1.28 and 1.27 in positive and negative
ion detect mode respectively. In addition, as revealed of fold
change (FC) in Tables 13, we found that LPCs were up-regulated
and PCs and TGs were down-regulated, compared C group with B
and N groups.
LPCs consist of fatty acids of varying lengths and degrees of
saturation attached at the C-1 (sn-1) position. Fatty acids con-
taining 16, 18, and 20 carbons are the most common. In blood
plasma, signicant amounts of LPC are formed by the enzyme,
lecithin:cholesterolacyltransferase (LCAT). The enzyme catalyzes
the transfer of fatty acids of position sn-2 of PC to the free cho-
lesterol in plasma, with subsequent formation of cholesterol esters
and LPC. Also LPC can be continuously produced by the activity of
phospholipase A2 (PLA2) through hydrolysis of PC. Recently, sev-
eral papers reported that the PLA2 activity is higher in OC tumor
tissue than benign tumor and normal tissue [26] and PLA2 activity
was suggested as a potential marker and therapeutic target for OC
 Y. Zhang et al. / Talanta 150 (2016) 8896 93

Fig. 4. PLS-DA score plots (A and C) and loadings-plots (B and D) based on the data from UPLC-ESI( )-QTOF-MS (A and B) or UPLC-ESI(  )-QTOF-MS (C and D) for
distinguishing C, B and N group, and for selecting the potential markers. The variables with VIP value 45.0 were highlighted with red boxes. (For interpretation of the
references to color in this gure legend, the reader is referred to the web version of this article.)

A 184.0740
B
30000
800000 305.2486

25000

600000
Intensity,cps

Intensity,cps

20000
768.5550

400000 15000
255.2328

10000
828.5766
200000
784.5856 806.5667 5000
766.5753

0 0
0 200 400 600 800 1000 1200 1400 0 200 400 600 800 1000 1200

m/z m/z

Fig. 5. MS/MS data of a variable with retention time (RT) at 5.72 min in positive-ion detect mode (A) and negative ion detect mode (B).

[27]. LPC molecules are involved in multiple biological processes, esophageal squamous cell carcinoma [29] and hepatocellular car-
including cell signaling, energy metabolism and storage, and cell cinoma [30]. Our result showed LPCs were upregulated and PCs
membrane integrity/stability. LPC levels have been shown to be were downregulated, which may be the result of higher PLA2 ac-
altered in a number of cancers, including breast cancer [28], tivity in OC patients. Moreover, TGs spillover from adipose tissue
94 Y. Zhang et al. / Talanta 150 (2016) 8896

A 700000
B 500000
872.7671
[M +H-R COOH]+:NL of fatty acid 18:2
1 1
600000 [M +H-R COOH]+:NL of fatty acid 16:0
877.7273 1 2
400000
500000 [M +H-R COOH]+:NL of fatty acid 18:1

Intensity,cps
2 1
Intensity,cps

[M +H-R COOH]+:NL of fatty acid 18:2

2 2
300000 575.5032
400000

300000 599.5023
200000 601.5072
898.7820
200000 903.7374 872.7671
893.6942 898.7820
100000
100000

0 0
800 820 840 860 880 900 920 940 600 800
m/z m/z

Fig. 6. MS data (A) and MS/MS data (B) of 2 variables with RT at around 15.04 min in positive ion detect mode.

Table 1
Signicant changed lipids compared C with N group obtained by positive and negative ion detect mode.

Potential biomarkers RT (min) Quasimolecular ion m/z Detected m/z Exact Error (ppm) p value FC (C/N)

14:0 LPC 0.58 [M HCOO]  512.3018 512.2994 4.68 0.003 1.16

16:0 LPC 1.12 [M H] 496.3414 496.3398 3.22 0.014 1.08
18:1 LPC 1.17 [M H] 522.3559 522.3554 0.96 0.019 1.28
[M HCOO]  566.3464 566.3463 0.18 0.002 1.27
20:3 LPC 1.06 [M HCOO]  590.3456 590.3463  1.19 0.004 1.19
20:4 LPC 0.94 [M H] 544.3402 544.3398 0.73 0.007 1.24
[M HCOO]  588.3304 588.3307  0.51 0.006 1.23
22:6 LPC 0.90 [M H] 568.3399 568.3398 0.18 0.026 1.22
16:0/18:1 PC 6.86 [M H] 760.5844 760.5851  0.92 0.008 0.54
[M HCOO]  804.5748 804.5760  1.49 0.008 0.56
16:0/18:2 PC 5.58 [M H] 758.5718 758.5694 3.16 0.021 0.78
16:0/20:3 PC 5.73 [M H] 784.5838 784.5851  1.66 0.027 0.68
[M HCOO]  828.5748 828.5760  1.45 0.003 0.61
16:0/20:4 PC 4.69 [M HCOO]  826.5587 826.5604  2.06 0.012 0.77
18:0/18:1 PC 9.00 [M HCOO]  832.6061 832.6073  1.44 0.023 0.72
18:0/18:2 PC 7.31 [M H] 786.6001 786.6007  0.76 0.013 0.54
18:0/20:5 PC 7.31 [M H] 808.5809 808.5851  5.19 0.041 0.93
18:2/18:2 PC 4.69 [M H] 782.5677 782.5694  2.17 3.1E  5 0.31
18:2/18:2/16:0 TG 15.04 [M NH4] 872.7671 872.7702  3.44 0.018 0.64

Table 2
Signicant changed lipids compared C and B group obtained by positive and negative ion detect mode.

Potential biomarkers RT (min) Quasimolecular ion m/z Detected m/z Exact Error (ppm) p value FC (C/B)

16:0 LPC 1.12 [M H] 496.3414 496.3398 3.22 0.002 1.14

[M HCOO]  540.3311 540.3307 0.74 0.0016 1.18
18:0 LPC 1.51 [M H] 524.3716 524.3712 0.76 0.009 1.19
[M HCOO]  568.3620 568.3620 0 0.0081 1.21
18:1 LPC 1.17 [M H] 522.3559 522.3554 0.96 0.005 1.23
[M HCOO]  566.3464 566.3463 0.18 0.0050 1.27
18:2 LPC 0.97 [M H] 520.3419 520.3398 4.04 0.0004 1.24
[M HCOO]  564.3310 564.3307 0.53 0.0003 1.23
20:3 LPC 1.06 [M HCOO]  590.3456 590.3463  1.19 0.001 1.36
20:4 LPC 0.94 [M H] 544.3402 544.3398 0.73 0.009 1.24
[M HCOO]  588.3304 588.3307  0.51 0.008 1.25
22:6 LPC 0.90 [M HCOO]  568.3399 568.3398 0.18 0.05 1.16
16:0/18:1 PC 6.84 [M HCOO]  804.5748 804.5760  1.49 0.025 0.67
16:0/18:2 PC 5.58 [M H] 758.5718 758.5694 3.16 0.01 0.86
18:0/18:2 PC 7.31 [M H] 786.6001 786.6007  0.76 0.016 0.60
18:0/20:5 PC 7.31 [M H] 808.5809 808.5851  5.19 0.034 0.95
18:2/18:2 PC 4.69 [M H] 782.5677 782.5694  2.17 0.0005 0.43
18:1/16:0 SM 4.99 [M H] 703.5739 703.5748  1.28 0.037 1.04
18:2/18:2/18:1 TG 15.05 [M NH4] 898.7817 898.7858  4.56 0.005 0.47
18:2/18:1/18:1 TG 15.32 [M NH4] 900.7972 900.8015  4.77 0.016 0.67
18:2/18:2/16:0 TG 15.04 [M NH4] 872.7671 872.7702  3.44 6.8E  5 0.50
18:1/18:2/16:0 TG 15.31 [M NH4] 874.7830 874.7858  3.20 2.4E  6 0.47
18:1/18:1/16:0 TG 15.57 [M NH4] 876.7984 876.8015  3.54 0.0001 0.47
 Y. Zhang et al. / Talanta 150 (2016) 8896 95

during periods of caloric excess leads to lipid accumulation in lipids were analyzed using pathway enrichment analysis by MetPA
nonadipose tissues, resulting in cellular dysfunction known as li- (Metabolomic Pathway Analysis) software. In our analysis, we
potoxicity [31]. Interestingly, our results showed a unique trend in identied several pathways, including glycerophospholipid meta-
PC, LPC and TG changes, but more work is required to fully char- bolism, linoleic acid metabolism, sphingolipid metabolism, alpha-
acterize this observation. Studying the cellular processes and ge- linolenic acid metabolism, glycerolipid metabolism and arachi-
netic components involved in the progression of these disorders is donic acid metabolism, that may be perturbed in OC. The sum-
necessary for the development of new treatment strategies. mary of pathway analysis is shown in Fig. 7. The detailed results of
pathway analysis are shown in Table 4.
As we all know, PCs and LPCs mentioned here belong to the
3.4. Pathway analysis class of glycerophospholipids; TGs belong to glycerolipids. Gly-
cerophospholipids are components of the cell membrane, and
The lipid metabolism was regulated by enzymes. To identify previous research has shown that highly proliferating cancer cells
pathways that are disregulated in patients with OC, the identied require continuous de novo fatty acid synthesis [31]. We hy-
pothesize that the altered glycerophospholipid metabolism and
Table 3 glycerolipid metabolism identied in our study reects abnormal
Potential biomarkers for differentiation of C, B and N group obtained by positive
proliferation of these cells. The mechanisms underlying this per-
and negative ion detect mode.
turbation and the enzymes and genes involved in this process
Potential RT (min) Quasimolecular ion m/z Detected m/z Exact Error require further study. Targeting of glycerophospholipid metabo-
biomarkers (ppm) lism is a therapeutic approach that has been applied to several
diseases [31]; however, it has not been tested as a treatment op-
16:0 LPC 1.12 [MH] 496.3414 496.3398 3.22
tion for OC. The work we present here provides a new link among
18:1 LPC 1.17 [MH] 522.3559 522.3554 0.96
[MHCOO]  566.3464 566.3463 0.18 glycerophospholipid metabolism, glycerolipid metabolism and OC.
20:3 LPC 1.06 [MHCOO]  590.3456 590.3463  1.19 This metabolic disturbance may play a key role in the pathogenesis
20:4 LPC 0.94 [MH] 544.3402 544.3398 0.73 of OC. Additionally, the metabolites may be useful biomarkers to
[MHCOO]  588.3304 588.3307  0.51
distinguish patients with OC from healthy individuals, and provide
22:6 LPC 0.90 [MH] 568.3399 568.3398 0.18
16:0/18:1 6.84 [MHCOO]  804.5748 804.5760  1.49 a new means of diagnosis.
PC
16:0/18:2 5.58 [MH] 758.5718 758.5694 3.16
PC 4. Conclusion
18:0/18:2 7.31 [MH] 786.6001 786.6007  0.76
PC
18:0/20:5 7.31 [MH] 808.5809 808.5851  5.19 This is the rst report using lipidomics to study OC in human
PC patients. A rapid, reliable, and sensitive UPLC-ESI-QTOF-MS
18:2/18:2 4.69 [MH] 782.5677 782.5694  2.17 method was validated for qualitative and relatively quantitative
PC
analysis of OC plasma samples. Patients with OC could be dis-
18:2/18:2/ 15.04 [MNH4] 872.7671 872.7702  3.44
16:0 TG tinguished from healthy individuals and patients with benign
gynecological tumor disease by this method, which shows a sig-
nicant lipid perturbation in those diseases. With the assistance of
high resolution and high accuracy of MS and MS/MS data, the
potential markers, including lysophosphatidylcholines (LPCs),
phosphatidylcholines (PCs) and triacylglycerols (TGs) with specic
fatty acid chains, were identied. Interestingly, LPCs were up-
regulated and PCs and TGs were down-regulated, compared OC
group with benign tumor and normal control groups. LPC can be
continuously produced by the activity of phospholipase A2 (PLA2)
through hydrolysis of PC. Recently, several papers reported that
the PLA2 activity is higher in OC tumor tissue than benign tumor
and normal tissue, and combined with our data (up-regulated LPCs
and down-regulated PCs), PLA2 activity was suggested as a po-
tential marker and therapeutic target for OC. The pathway analysis
also showed that the glycerophospholipid metabolism emerged as
a key pathway that was disregulated in the disease. This study may
provide new insight into underlying mechanisms for OC and
proves that MS technology combined with lipidomics is a powerful
method in discovering new potential clinical biomarkers for
Fig. 7. Summary of pathway analysis. diseases.

Table 4
Results from pathway analysis.

Total Expected Hits Raw p  log(p) Holm adjust FDR Impact

Glycerophospholipid metabolism 39 0.06 2 1.50E  03 6.50E 00 1.20E  01 1.20E  01 0.10

Linoleic acid metabolism 15 0.02 1 2.47E  02 3.70E 00 1.00E 00 8.35E  01 0.00
Sphingolipid metabolism 25 0.04 1 4.09E  02 3.20E 00 1.00E 00 8.35E  01 0.01
alpha-Linolenic acid metabolism 29 0.05 1 4.74E  02 3.05E 00 1.00E 00 8.35E  01 0.00
Glycerolipid metabolism 32 0.05 1 5.22E  02 2.95E 00 1.00E 00 8.35E  01 0.05
Arachidonic acid metabolism 62 0.10 1 9.92E  02 2.31E 00 1.00E 00 1.00E 00 0.00
96 Y. Zhang et al. / Talanta 150 (2016) 8896
Notes
The authors declare no competing nancial interest.
Acknowledgments
This work was supported by the Science and Technology Pro-
gram of Beijing Municipality (No. Z 131100005213009) and Na-
tional Natural Science Foundation of China (Grant nos. 21321003
and 21405160).
References
[1] N. Howlader, A.N. Noone, M. Krapcho, J. Garshell, D. Miller, S.F. Altekruse, C.
L. Kosary, M. Yu, J. Ruhl, Z. Tatalovich, A. Mariotto, D.R. Lewis, H.S. Chen, E.
J. Feuer, K.A. Cronin, Seer Cancer Stat. Rev. (2014) 19752011.
[2] D. Badgwell Jr., R.C. Bast, Dis. Markers 23 (2007) 397410.
[3] M.M. Fields, E. Chevlen, Clin. J. Oncol. Nurs. 10 (2006) 7781.
[4] N.K. Wong, R.L. Easton, M. Panico, M. Sutton-Smith, J.C. Morrison, F.
A. Lattanzio, H.R. Morris, G.F. Clark, A. Dell, M.S. Patankar, J. Biol .Chem. 278
(2003) 2861928634.
[5] E.A. Eisenhauer, J.B. Vermorken, M. van Glabbeke, Ann. Oncol. 8 (1997)
963968.
[6] G.J. Rustin, Eur. J. Cancer 28 (1992) 23.
[7] G.J. Rustin, A.E. Nelstrop, P. McClean, M.F. Brady, W.P. McGuire, W.J. Hoskins,
H. Mitchell, H.E. Lambert, J. Clin. Oncol. 14 (1996) 15451551.
[8] R.G. Moore, M. Jabre-Raughley, A.K. Brown, K.M. Robison, M.C. Miller, W.
J. Allard, R.J. Kurman, R.C. Bast, S.J. Skates, Am. J. Obs. Gynecol. 203 (2010)
e1e6.
[9] R.G. Moore, D.S. McMeekin, A.K. Brown, P. DiSilvestro, M.C. Miller, W.J. Allard,
W. Gajewski, R. Kurman, R.C. Bast, S.J. Skates Jr., Gynecol. Oncol. 112 (2009)
4046.
[10] R.G. Moore, M.C. Miller, P. Disilvestro, L.M. Landrum, W. Gajewski, J.J. Ball, S.
J. Skates, Obstet. Gynecol. 118 (2011) 280288.
[11] H.A. Hirsch, D. Iliopoulos, A. Joshi, Y. Zhang, S.A. Jaeger, M. Bulyk, P.N. Tsich-lis,
S.X. Liu, K. Struhl, Cancer Cell 17 (2010) 348361.
[12] Y.J. Xiao, B. Schwartz, M. Washington, A. Kennedy, K. Webster, J. Belinson,
Y. Xu, Anal. Biochem. 290 (2001) 302313.
[13] Y. Xu, Z.Z. Shen, D.W. Wiper, M.Z. Wu, R.E. Morton, P. Elson, A.W. Kennedy,
J. Belinson, M. Markman, G. Casey, JAMA 280 (1998) 719723.
[14] M. Meleh, B. Pozlep, A. Mlakar, H. Meden-Vrtovec, L. Zupancic-Kralj, J. Chro-
matogr. B 858 (2007) 287291.
[15] Z. Zhao, Y. Xiao, P. Elson, H. Tan, S.J. Plummer, M. Berk, P.P. Aung, I.C. Lavery, J.
P. Achkar, L. Li, et al., J. Clin. Oncol. 25 (2007) 26962701.
[16] A.D. Patterson, O. Maurhofer, D. Beyoglu, et al., Cancer Res. 71 (2011)
65906600.
[17] F. Li, X.Z. Qin, H.Q. Chen, L. Qiu, Y.M. Guo, H. Liu, G.Q. Chen, G.G. Song, X.
D. Wang, F.J. Li, S. Guo, B.H. Wang, Z.L. Li, Rapid Commun. Mass. Spectrom. 27
(2013) 2434.
[18] X.L. Han, S. Rozen, S.H. Boyle, C. Hellegers, H. Cheng, J.R. Burke, K.A. Welsh-
Bohmer, P.M. Doraiswamy, R. Kaddurah-Daouk, Plos ONE 6 (2011) e21643.
[19] Y.B. Wei, S.M. Li, J.X. Wang, C.Y. Shu, J.A. Liu, S.X. Xiong, J.W. Song, J.J. Zhang, Z.
W. Zhao, Anal. Chem. 85 (2013) 47294734.
[20] L. Lin, L.L. Wang, D.H. Shangguan, Y.B. Wei, J.J. Han, S.X. Xiong, Z.W. Zhao, J.
Chromatogr. A 1381 (2015) 140148.
[21] Y.B. Wei, Y.Y. Zhang, Y. Lin, L. Li, J.A. Liu, Z.P. Wang, S.X. Xiong, Z.W. Zhao,
Analyst 140 (2015) 12981305.
[22] L. Li, Y.Y. Zhang, Z.W. Zhao, Sci. Sin. Chim. 44 (2014) 17.
[23] A. Fauland, H. Kfeler, M. Trtzmller, A. Knopf, J. Hartler, A. Eberl, C. Chitraju,
E. Lankmayr, F. Spener, J. Lipid Res. 52 (2011) 23142322.
[24] M.R. Wenk, Cell 143 (2010) 888895.
[25] Z.W. Zhao, Y. Xu, J. Lipid Res. 51 (2010) 652659.
[26] H. Li, Z.W. Zhao, G. Wei, L.B. Yan, D.M. Wang, H. Zhang, G.E. Sandusky, J. Turk,
Y. Xu, FASEB J. 26 (2012) 33063320.
[27] H. Li, Z.W. Zhao, C. Antalis, Z.Z. Zhao, R. Emerson, G. Wei, S. Zhang, Z.Y. Zhang,
Y. Xu, Am. J. Pathol. 179 (2011) 452461.
[28] Y.P. Qiu, B.S. Zhou, M. Su, S. Baxter, X.J. Zheng, X.Q. Zhao, Y. Yen, W. Jia, Int. J.
Mol. Sci. 14 (2013) 80478061.
[29] J. Xu, Y. Chen, R. Zhang, Y. Song, J. Cao, N. Bi, J. Wang, J. He, J. Bai, L. Dong,
L. Wang, Q. Zhan, Z. Abliz, Mol. Cell. Proteom. 12 (2013) 13061318.
[30] H.W. Ressom, J.F. Xiao, L. Tuli, R.S. Varghese, B. Zhou, T.H. Tsai, M.R.N. Ranjbar,
Y. Zhao, J.L. Wang, C.D. Poto, A.K. Cheema, M.G. Tadesse, R. Goldman, K. Shetty,
Anal. Chim. Acta 743 (2012) 90100.
[31] K.V. Ruggles, A. Turkish, S.L. Sturley, Annu. Rev. Nutr. 33 (2013) 413451.