Discovery and identification of potential biomarkers of papillary thyroid carcinoma.

Abstract

Thyroid carcinoma is the most common endocrine malignancy and a common cancer among the malignancies of head and neck. Noninvasive and convenient biomarkers for diagnosis of papillary thyroid carcinoma (PTC) as early as possible remain an urgent need. The aim of this study was to discover and identify potential protein biomarkers for PTC specifically.
Two hundred and twenty four (224) serum samples with 108 PTC and 116 controls were randomly divided into a training set and a blind testing set. Serum proteomic profiles were analyzed using SELDI-TOF-MS. Candidate biomarkers were purified by HPLC, identified by LC-MS/MS and validated using ProteinChip immunoassays.
A total of 3 peaks (m/z with 9190, 6631 and 8697 Da) were screened out by support vector machine (SVM) to construct the classification model with high discriminatory power in the training set. The sensitivity and specificity of the model were 95.15% and 93.97% respectively in the blind testing set. The candidate biomarker with m/z of 9190 Da was found to be up-regulated in PTC patients, and was identified as haptoglobin alpha-1 chain. Another two candidate biomarkers (6631, 8697 Da) were found down-regulated in PTC and identified as apolipoprotein C-I and apolipoprotein C-III, respectively. In addition, the level of haptoglobin alpha-1 chain (9190 Da) progressively increased with the clinical stage I, II, III and IV, and the expression of apolipoprotein C-I and apolipoprotein C-III (6631, 8697 Da) gradually decreased in higher stages.
We have identified a set of biomarkers that could discriminate PTC from non-cancer controls. An efficient strategy, including SELDI-TOF-MS analysis, HPLC purification, MALDI-TOF-MS trace and LC-MS/MS identification, has been proved successful.

Full text

Page 1

BioMed Central
Page 1 of 14
(page number not for citation purposes)
Molecular Cancer
Open Access
Research
Discovery and identification of potential biomarkers of papillary
thyroid carcinoma
Yuxia Fan1, Linan Shi2, Qiuliang Liu1, Rui Dong1, Qian Zhang1,
Shaobo Yang1, Yingzhong Fan1, Heying Yang1, Peng Wu2, Jiekai Yu3,
Shu Zheng3, Fuquan Yang*2 and Jiaxiang Wang*1
Address: 1Department of Surgery, the First Affiliated Hospital, Zhengzhou University, Zhengzhou, Henan 450052, PR China, 2Proteomic Platform,
Institute of Biophysics, Chinese Academy of Sciences, Beijing 100101, PR China and 3Institute of Cancer, the Second Affiliated Hospital, College
of Medicine, Zhejiang University, Hangzhou, Zhejiang 310009, PR China
Email: Yuxia Fan - fanyuxia1981@126.com; Linan Shi - lianli_shi@126.com; Qiuliang Liu - qiuliang_liu@126.com;
Rui Dong - dongrui_good@126.com; Qian Zhang - zhang_qian2006@163.com; Shaobo Yang - shao_bo_yang@126.com;
Yingzhong Fan - yingzhong_fan@126.com; Heying Yang - heying_yang@163.com; Peng Wu - peng_wu008@yahoo.cn;
Jiekai Yu - jiekai_yu@yahoo.cn; Shu Zheng - shu_zheng@yahoo.cn; Fuquan Yang* - fuquan_yang@yahoo.cn;
Jiaxiang Wang* - wjiaxiang@zzu.edu.cn
* Corresponding authors
Abstract
Background: Thyroid carcinoma is the most common endocrine malignancy and a common
cancer among the malignancies of head and neck. Noninvasive and convenient biomarkers for
diagnosis of papillary thyroid carcinoma (PTC) as early as possible remain an urgent need. The aim
of this study was to discover and identify potential protein biomarkers for PTC specifically.
Methods: Two hundred and twenty four (224) serum samples with 108 PTC and 116 controls
were randomly divided into a training set and a blind testing set. Serum proteomic profiles were
analyzed using SELDI-TOF-MS. Candidate biomarkers were purified by HPLC, identified by LC-MS/
MS and validated using ProteinChip immunoassays.
Results: A total of 3 peaks (m/z with 9190, 6631 and 8697 Da) were screened out by support
vector machine (SVM) to construct the classification model with high discriminatory power in the
training set. The sensitivity and specificity of the model were 95.15% and 93.97% respectively in the
blind testing set. The candidate biomarker with m/z of 9190 Da was found to be up-regulated in
PTC patients, and was identified as haptoglobin alpha-1 chain. Another two candidate biomarkers
(6631, 8697 Da) were found down-regulated in PTC and identified as apolipoprotein C-I and
apolipoprotein C-III, respectively. In addition, the level of haptoglobin alpha-1 chain (9190 Da)
progressively increased with the clinical stage I, II, III and IV, and the expression of apolipoprotein
C-I and apolipoprotein C-III (6631, 8697 Da) gradually decreased in higher stages.
Conclusion: We have identified a set of biomarkers that could discriminate PTC from non-cancer
controls. An efficient strategy, including SELDI-TOF-MS analysis, HPLC purification, MALDI-TOF-
MS trace and LC-MS/MS identification, has been proved successful.
Published: 28 September 2009
Molecular Cancer 2009, 8:79 doi:10.1186/1476-4598-8-79
Received: 1 March 2009
Accepted: 28 September 2009
This article is available from: http://www.molecular-cancer.com/content/8/1/79
© 2009 Fan et al; licensee BioMed Central Ltd.
This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0),
which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

Page 2

Molecular Cancer 2009, 8:79 http://www.molecular-cancer.com/content/8/1/79
Page 2 of 14
(page number not for citation purposes)
Background
Thyroid carcinoma is the most common endocrine malig-
nancy and a common cancer among the malignancies of
head and neck. It comprises 91.5% of all endocrine malig-
nancies and 1% of all malignant diseases [1]. An esti-
mated 33550 new cases are diagnosed annually in the
United States and recent statistics shows the incidence of
thyroid carcinoma has increased, especially in papillary
thyroid carcinomas (PTC) [2]. PTC is the most common
type, which accounts for 80% of all thyroid cancers [3].
Early accurate diagnosis and timely treatment are critical
for improving long-term survival of PTC patients. Many
diagnostic tools have been used for thyroid carcinoma,
such as sonography, computed tomography, magnetic
resonance imaging, cytological examination and fine-nee-
dle aspiration. Currently, although ultrasound-guided
fine-needle aspiration biopsy is considered as the most
effective test for distinguishing malignant from benign
thyroid nodules, its sensitivity is approximately 93% and
its specificity is 75% [4]. At the same time, researchers
have been seeking valuable biomarkers for thyroid carci-
noma diagnosis, such as galectin-3, fibronectin-1, CITED-
1, HBME1, cytokeratin-19 and TPO, and so on. What is
disappointing is that all these biomarkers either are lack-
ing specificity to some degree, or have a poor positive pre-
dictive value [5-9]. To distinguish a malignant thyroid
nodule from a benign lesion more accurately, the diag-
nostic test, however, still needs to be improved. Moreover,
a noninvasive screening method for thyroid malignancy
remains unavailable.
Recent advances in the proteomics study have introduced
novel techniques for the screening of cancer biomarkers
and improved early and accurate diagnosis of cancer dis-
eases to a new horizon [10]. Surfaced enhanced laser des-
orption/ionization time of flight mass spectroscopy
(SELDI-TOF-MS), which generates the protein fingerprint
by MS, has been proved a powerful tool for potential
biomarker discovery [11,12]. Recently, the SELDI-TOF-
MS analysis has been successfully used to identify specific
biomarkers for various cancers, such as ovarian cancer,
prostate cancer, pancreatic cancer, colon cancer, breast
cancer, etc [13-17]. In search of biomarkers for diagnosing
PTC, a few pilot studies based on proteomics were con-
ducted, in which SELDI-TOF-MS has been utilized
[18,19]. However, no specific protein biomarkers have
been identified and validated in those reports.
In this study, firstly, we used SELDI-TOF-MS technology
to screen potential protein patterns specific for PTC and
then purified the candidate protein biomarker peaks by
HPLC, identified by LC-MS/MS and finally confirmed
these biomarkers by ProteinChip Immunoassays. To the
best of our knowledge, this is the first time that proteins
biomarkers have been identified for PTC.
Results
Serum protein profiles and data processing
Serum samples from the training set were analyzed and
compared by SELDI-TOF-MS with WCX2 chip. All MS
data were baseline subtracted and normalized using total
ion current, and the peak clusters were generated by
Biomarker Wizard software. After carrying out Wilcoxon
rank sum tests to determine relative signal strength, 26
peaks with p value < 0.01 were obtained. Seven protein
peaks were found up-regulated and 19 peaks were found
down-regulated in PTC group (data not shown). From the
random combination of protein peaks with remarkable
variation, support vector machine (SVM) screened out the
combined model with maximum Youden index of the
predicted value, identifying 3 markers positioned at 9190,
6631 and 8697 respectively. In the PTC group, the 9190
Da protein was remarkably elevated while 6631 & 8697
Da proteins were significantly decreased (figure 1). The
descriptive statistics of these 3 markers are shown in Table
1. In addition, the level of 9190 Da protein progressively
increased with the clinical stage I, II, III and IV, and the
expression of 6631, 8697 Da proteins gradually decreased
in higher stages (figure 2). Combining 3 potential mark-
ers, using the method of leave-1-out for cross detection,
the sensitivity of discriminating 60 PTC and 40 normal
subjects was 98%, and its specificity was 97%.
Protein peak validation
The remaining 48 PTC and 76 control serum samples (20
healthy controls and 56 patients with benign thyroid
node) as a blind testing set, were analyzed to validate the
accuracy and validity of the classification model derived
from the training set. The descriptive statistics of the three
markers in 48 PTC patients and 76 non-cancer controls
are shown in Table 2. The classification model distin-
guished the PTC samples from controls with a sensitivity
of 95.15%, specificity of 93.97%, and positive predictive
value of 96.0%, respectively. The area under the receiver
operating characteristics (ROC) curve of this model was
0.971.
Purification and identification of candidate protein
biomarkers
Serum samples from PTC patients were used for the puri-
fication of the up-regulated candidate protein biomarker
(9190 Da), and serum samples from healthy controls
Table 1: The relative peak intensity of three distinct protein
spectra found in sera of PTC patients.
m/z
Patients with PTC
(mean ± SD)
Healthy Individuals
(mean ± SD)
p
9190
6631
8697
6853.82 ± 1585.23
2706.56 ± 578.17
3017.98 ± 600.28
282.46 ± 118. 09
4697.16 ± 989.85
4924.32 ± 1048.11
0.000538
0.001381
0.001672

Page 3

Molecular Cancer 2009, 8:79http://www.molecular-cancer.com/content/8/1/79
Page 3 of 14
(page number not for citation purposes)
A representative mapping of SELDI-TOF-MS analysis of sera from PTC patients and healthy controls
Figure 1
A representative mapping of SELDI-TOF-MS analysis of sera from PTC patients and healthy controls. Differen-
tially expressed proteins with potential diagnostic significance are arrowed. Top-group denotes sera from patients with PTC, in
which the protein with m/z of 9190 Da was over-expressed. Bottom-group denotes sera from healthy individuals, in which the
proteins with m/z of 6631 and 8697 Da were up-regulated.

Page 4

Molecular Cancer 2009, 8:79http://www.molecular-cancer.com/content/8/1/79
Page 4 of 14
(page number not for citation purposes)
A representative mapping of SELDI-TOF-MS analysis from different stages of PTC patients and non-cancer controls
Figure 2
A representative mapping of SELDI-TOF-MS analysis from different stages of PTC patients and non-cancer
controls. The level of the 9190 Da protein progressively increased with the clinical stage I, II, III and IV, and the expression of
6631 and 8697 Da proteins gradually decreased in higher stages.

Page 5

Molecular Cancer 2009, 8:79http://www.molecular-cancer.com/content/8/1/79
Page 5 of 14
(page number not for citation purposes)
were used for the purification of the two down-regulate
proteins (6631, 8697 Da) using WCX SPE and C18 HPLC.
Figure 3 shows the results of MALDI-TOF-MS analysis of
the three purified candidate protein biomarkers.
After digestion with modified trypsin, the peptide mixture
was analyzed by nano-LC-MS/MS. Figure 4 shows the
results of the LC-MS/MS chromatogram (A) and MS/MS
spectrum of one identified peptide (B) from protein
(8697 Da). Table 3 shows the results of identification of
the three candidate protein biomarkers. They were hap-
toglobin alpha-1 chain (9190 Da) [NCBI:P00738], apoli-
poprotein C-I (6631 Da)
apolipoprotein C-III (8697 Da) [NCBI:CAA25233]. The
whole sequence of the three candidate protein markers is
given by combination of high sequence coverage and
accurate molecular weight (MW) measurement using
MALDI-TOF-MS.
[NCBI:P02654] and
Validation of three candidate protein biomarkers
A ProteinChip-array-based immunoassay (Ciphergen Bio-
systems) was used to specifically capture haptoglobin
alpha-1 chain, apolipoprotein C-I and apolipoprotein C-
III from crude serum samples and to confirm the signifi-
cance of each marker. The anti-haptoglobin alpha-chain
antibody specifically captured the previously identified
9190 Da protein. The anti-apolipoprotein C-I array was
developed to capture apolipoprotein C-I (6631 Da) and
the apolipoprotein C-III antibody against specifically cap-
tured apolipoprotein C-III (8697 Da). (Figure 5)
Discussions
In this study, we obtained serum protein mass spectra
from PTC patients and controls using SELDI-TOF-MS.
Based on the serum proteomic profiles, we constructed a
classification model to discriminate PTC patients from
non-cancer controls. One of the challenges in the analysis
of SELDI-TOF-MS-generated data is to reduce the false
protein peaks, in which the discriminatory power is due to
random variation [20]. To solve this problem in the data
processing of this experiment, we eliminated noise by dis-
crete wavelength, identified mass-charge peaks of speci-
mens using the method of local extremum, and clustered
mass-charge peaks by setting 10% as the minimum
threshold. Wilcoxon rank sum test analysis assessed the
relative importance of each peak in the discrimination of
2 kinds of specimen according to P values. Furthermore,
SVM was employed in our experiment, which is a kind of
classification technology proposed by Vapnik and others.
In the model discrimination, the popularization, model
selection, overfitting, latitude disaster, and other prob-
lems of the small specimen model have been solved suc-
cessfully in SVM [21-23]. The procedures included
randomly combining the remarkably different mass-
charge peaks and inputting them into SVM, screening out
the markers, building the discrimination model, and then
using the method of leave one out to assess the model by
means of cross verification. By combination among these
procedures mentioned above, the popularization of the
model building and the accuracy of the prediction were
ensured. The classification model could discriminate
patients with PTC from non-cancer controls with a sensi-
tivity of 95.15% and a specificity of 93.97% in the blind
testing set. The up-regulated candidate protein biomarker
was identified as haptoglobin alpha-1 chain (9190 Da).
Another two down-regulated candidate protein biomark-
ers (6631 and 8697 Da) were identified as apolipoprotein
C-I and apolipoprotein C-III.
Among the proteins identified by LC-MS/MS, hap-
toglobin alpha-1 chain was significantly elevated in PTC
patients and this protein may play a critical role in the
development of PTC. Intact haptoglobin, composed of
two different polypeptides (alpha and beta-chains), is an
acute phase protein capable of binding haemoglobin and
preventing iron loss [24]. It was reported that body iron
could promote neoplastic cell growth and accumulate in
cancer cells more than in normal cells [25]. Furthermore,
a few other studies have demonstrated that there is a
higher cancer risk in patients with larger iron stores than
those with small iron stores [26]. Collectively, highly
expressed haptoglobin results in a high hemoglobin/iron
existence and raises the possibility of a causative involve-
ment of iron-derived oxidative stress in the tumour devel-
opment. Recently, there are several reports in the literature
showing increased expression of haptoglobin in ovarian
cancer, prostatic carcinoma and pancreatic cancer [27-29],
and the level of haptoglobin alpha chain up-regulated in
serum of breast cancer [30]. However, it is necessary to
understand more clearly of the role of haptoglobin alpha-
1 chain in the development of PTC through further stud-
ies of biological mechanisms of thyroid carcinoma.
Apolipoproteins (APOs) are lipid carriers and previous
studies about APO mainly focused on lipoprotein metab-
olism. Recently, APOs have been reported to regulate
many cellular function. For example, the protein kinase
Akt can be elicited by APO C-I, which in turn promotes
growth factor-mediated cell survival and block apoptosis
[31]. In this study, the APO C-I is down-regulated in the
serum of PTC patients, which indicates that APO C-I may
be related to PTC. Thus, besides the function of APO C-I
in lipid metabolism, additional function of APO C-I in
Table 2: The descriptive statistics of these 3 markers in the blind
testing set.
m/z
Patients with PTC
(mean ± SD)
Non-cancer Controls
(mean ± SD)
p
9190
6631
8697
6798.89 ± 1517.76
2786.03 ± 596.11
3011.62 ± 598.13
271.35 ± 102.04
4705.23 ± 991.45
4912.36 ± 1042.79
0.000526
0.001402
0.001669