Serum Protein Profiles in Myasthenia Gravis

Chao Cheng, Guoyong Wu, Sai-Ching J. Yeung, Rong Li, Amos Ela Bella, Jinzhuo
Pang, Fo-tian Zhong, Honghe Luo, Yanli Jin and Jingxuan Pan
Ann Thorac Surg 2009;88:1118-1123
DOI: 10.1016/j.athoracsur.2009.05.032

The online version of this article, along with updated information and services, is
located on the World Wide Web at:
http://ats.ctsnetjournals.org/cgi/content/full/88/4/1118

The Annals of Thoracic Surgery is the official journal of The Society of Thoracic Surgeons and the
Southern Thoracic Surgical Association. Copyright 2009 by The Society of Thoracic Surgeons.
Print ISSN: 0003-4975; eISSN: 1552-6259.

Downloaded from ats.ctsnetjournals.org by on February 10, 2012

GENERAL THORACIC
Serum Protein Profiles in Myasthenia Gravis
Chao Cheng, MD, PhD, Guoyong Wu, MD, Sai-Ching J. Yeung, MD, PhD,
Rong Li, MD, PhD, Amos Ela Bella, MS, Jinzhuo Pang, MD, Fo-tian Zhong, MD,
Honghe Luo, MD, Yanli Jin, MS, and Jingxuan Pan, MD, PhD
Department of Thoracic Surgery, The First Affiliated Hospital of Sun Yat-sen University, Lung Cancer Research Institute of
Guangdong Provincial Peoples Hospital, and Department of Pathophysiology, Zhongshan School of Medicine, Sun Yat-sen
University, Guangzhou, Peoples Republic of China; and Department of General Internal Medicine, Ambulatory Treatment and
Emergency Care, The University of Texas M. D. Anderson Cancer Center, Houston, Texas
Background. The diagnosis of myasthenia gravis (MG)
remains challenging. We performed a proteome-wide
search for potential serum protein diagnostic markers for
MG using surface-enhanced laser desorption/ionization
(SELDI) time-of-flight mass spectrometry (TOFMS).
Methods. Proteomic spectra from 80 MG patients and
80 healthy individuals were generated by SELDI. Sam-
ples from 56 MG patients and 56 healthy individuals in
the training set were analyzed to set up the decision tree.
Samples from 24 MG patients and 24 healthy individuals
were used for cross-validation testing.
Results. The SELDI TOFMS analysis generated 101 peaks,
representing differentially expressed proteins between 1000
and 20000 Da. Among them, 9 peaks were down-regulated
and 30 others were up-regulated in the MG sera compared
M yasthenia gravis (MG) is an autoimmune neuro-
muscular junction disorder that causes muscle
weakness. The incidence rate for MG has increased over
time [1], and this phenomenon is likely due to improve-
ment in diagnosis. The diagnosis for MG remains chal-
lenging because of its fluctuating clinical course and
signs and symptoms similar to other neuromuscular
disorders such as Lambert-Eaton syndrome, botulism,
congenital myasthenic syndromes, and tick paralysis [1].
At present, there are still no widely accepted diagnostic
criteria for MG [2].
Surface-enhanced laser desorption/ionization (SELDI)
time-of flight mass spectrometry (TOFMS) technology
combines protein chip array with TOFMS and offers not
only the advantages of speed, simplicity, sensitivity and
suitability for a comparative study but also the advantage
of using a very small amount of sample for the test. In
recent years, this technique has been successfully used to
study biomarkers of certain cancers [35]. In this study,
we report the application of SELDI TOFMS in the diag-
nosis of MG using serum samples.
Accepted for publication May 13, 2009.
Address correspondence to Dr Pan, Department of Pathophysiology,
Zhongshan School of Medicine, Sun Yat-sen University, Guangzhou,
510089, Peoples Republic of China; e-mail: jingx_pan@yahoo.com.cn.
2009 by The Society of Thoracic Surgeons
Published by Elsevier Inc
Downloaded from ats.ctsnetjournals.org by on February 10, 2012
with the controls. The decision tree used the peak at
M4168.94 Da and M1122.57 Da as splitters in the classifica-
tion process. In the training set, 112 samples were classified
as MG or control group, with a sensitivity of 100% and
specificity of 89.3%; the 10-fold cross-validated analysis
identified the optimal decision tree with the lowest relative
cross-validated cost of 0.080. In the test set, the decision tree
generated was able to identify 20 of 24 MG patients and 21
of 24 healthy individuals with a sensitivity of 83.3% and a
specificity of 87.5%.
Conclusions. SELDI TOFMS is a useful tool for the
detection and identification of potential serum biomarkers
that can diagnose MG with high sensitivity and specificity.
(Ann Thorac Surg 2009;88:1118 23)
2009 by The Society of Thoracic Surgeons
Material and Methods
This study was approved by the SunYat-sen University
Institutional Review Board. All study participants provided
informed consent according to institutional guidelines.
Patients
Peripheral blood samples were obtained before thymec-
tomy from 80 patients (34 males and 46 females) aged 4 to 48
years old who were diagnosed with MG during June 2007 to
June 2008 in The First Affiliated Hospital of Sun Yat-sen
University. The MG patients underwent extended thymec-
tomy consisting of complete en bloc extirpation of thymic
and adjacent tissues, including fatty tissue through median
sternotomy, followed by pathologic examination.
All patients met all of the five diagnostic criteria for
MG [1]:
1. pharmacologic testing with edrophonium chloride
that elicits unequivocal improvement in strength;
2. electrophysiologic testing with repetitive nerve
stimulation studies or single-fiber electromyogra-
phy (SFEMG), or both, demonstrating a primary
postsynaptic neuromuscular junctional disorder;
3. typical clinical manifestation demonstrating oculo-
motor weakness with asymmetric ptosis and binoc-
ular diplopia or asymmetrical weakness of multiple
extraocular muscles that cannot be attributed to
neuropathy involving a single cranial nerve;
0003-4975/09/$36.00
doi:10.1016/j.athoracsur.2009.05.032
Ann Thorac Surg
2009;88:1118 23
4. therapeutic benefits from anticholinesterase drugs
or corticosteroids, or both; and
5. pathologic findings of hyperplasia of the thymus.
According to the Osserman classification [6], 41 of 80
patients were in class I (ie, ocular MG); 16 were class IIa;
18 were class IIb; and 5 were class III. The peripheral
blood samples of 80 healthy individuals (36 males and 44
females) aged 13 to 40 years were used as the control.
Preparation of Specimens
All blood samples were allowed to clot at 4C and
centrifuged at 3000 rpm for 10 minutes. The serum was
collected and frozen in aliquots for storage at 80C.
SELDI Processing Of Serum Samples
Serum samples were processed robotically on a Ciphergen
Biosystems (Freemont, CA) liquid handling system in an
8-well format for SELDI analysis. A 10-L aliquot of serum
was pretreated with 9M urea, 2% 3-[(3-cholamidopropyl)-
dimethylammonio]-1-propane sulfonate (CHAPS), and
50mM Tris, and was vortexed for 30 minutes at 4C. Further
dilution was made in 50mM sodium acetate binding buffer.
Each sample position was spotted onto WCX-2 chips. Pro-
tein chips were used for SELDI TOFMS analysis with the
aid of a bioprocessor. The pretreated samples were added
to each well of chips and incubated for 30 minutes. After
washing three times with 200 L binding buffer (50mM
sodium acetate, pH 7.4) and deionized water, the chips were
removed from the bioprocessor and air-dried for 20
minutes.
A saturated solution of sinapinic acid (0.5 L) in 50%
acetonitrile and 0.5% trifluoroacetic acid was applied to
each spot on the chip surfaces. The chips were read on a
protein biologic system II (PBS-II) and a MS reader
(Ciphergen Biosystems Inc). The molecular mass per
charge (m/z) and relative intensity of each protein ad-
hered to the chip were measured with TOFMS. The mass
spectra were obtained by averaging 140 laser shots with
an intensity of 200, a detector sensitivity of 9, a peak mass
of 30,000 Da, and an optimized range of 1000 to 20,000 Da.
The mass accuracy was calibrated to less than 0.1% using
the All-in-one Peptide Molecular Mass Standard (Ci-
phergen Biosystems Inc). Peaks were autodetected when
occurring in at least 30% of spectra and with first- and
second-pass signal-to-noise (S/N) of 5 and 2, respec-
tively, in a 0.3% cluster mass window.
SELDI Data Analysis
Biomarker Wizard (BMW) software (Ciphergen Biosys-
tems) was used for peak clustering. Pattern recognition
and sample classification were performed with the Bi-
omarker Pattern Software (Ciphergen Biosystems), in
which multiple decision trees were initially generated
with the use of all the peaks as variables. A learning set
of 56 samples of MG patients and 56 healthy controls
were used to set up the decision tree. The 10-fold cross-
validated analysis was used to identify the optimal deci-
sion tree with the overall lowest relative cross-validated
cost. Under 10-fold cross validation analysis, all the
Downloaded from ats.ctsnetjournals.org by on February 10, 2012
CHENG ET AL 1119
GENERAL THORACIC
DIAGNOSING MG BY SERUM PROTEINS
computational work of generating trees is repeated at
least 10 times, which can assess the validation of the
performance of the decision tree.
Next, the decision tree was subjected to external cross-
validation (a blind test). A blinded validation set of 24
samples from MG patients and 24 samples from healthy
controls were categorized using the decision tree and
compared with their true MG status to evaluate the
validity and diagnostic performance of the decision tree.
The peaks that formed the main splitters of the tree with
the highest prediction rates in the training set was
selected and used to lay out a final decision tree with the
greatest possible predictive power.
Sensitivity is defined as the percentage of MG patients
in the validation set correctly detected by SELDI (true
positive/total number of MG). Specificity is defined as
the percentage of normal controls in the validation set
correctly identified by SELDI (true negative/total number
of normal controls). The positive likelihood ratio is de-
fined as the ratio of true positive rate/negative positive
rate. The negative likelihood ratio is defined as the ratio
of false negative rate/true negative rate. The positive
predictive value (PPV) is the proportion of patients with
positive test results who are correctly diagnosed. The
negative predictive value (NPV) is the proportion of
patients with negative test results who are correctly
diagnosed.
Positive likelihood ratio Sen (1 Spe);
negative likelihood ratio (1 Sen) Spe
PPV P0Sen [P0Sen (1 P0)(1 Spe)];
NPV (1 P0)(Spe) [(1 P0)Spe P0(1 Sen)]
With Sen, sensitivity; Spe, specificity; and P0, prevalence
in the population.
Results
In the training set, sera samples from 56 MG patients vs
the 56 healthy individuals were analyzed by affinity
Protein Chips. The resultant SELDI TOFMS data were
analyzed with Biomarker Wizard and generated 101
peaks representing differentially expressed proteins
ranged from 1000 to 20,000 Da. Of these, 39 protein peaks
had significant differences (p 005) between the MG and
control groups. Among them, 30 peaks were increased in
quantities (Table 1) and 9 peaks were decreased in
quantities (Table 2) in the MG sera compared with the
control sera. Figure 1 illustrates representative SELDI
spectra graphs with the differentially expressed peak
intensities between the MG group and the control group.
The expression of proteins at M3951.23, M3969.56,
M4168.94, and M4276.40 positions in MG sera were
reduced in comparison with those in normal control sera.
The peak intensity values of the 101 differentially
expressed peaks were collectively applied to the Biomar-
ker patterns algorithm program to generate an optimal
decision and classification tree. Under 10-fold cross val-
idation analysis, the output showed a tree sequence
summary, with 3 terminal nodes identified as optimal
 1120 CHENG ET AL Ann Thorac Surg
GENERAL THORACIC

DIAGNOSING MG BY SERUM PROTEINS 2009;88:1118 23

Table 1. Comparison of Protein Peaks That Were sensitivity of 83.3% and a specificity of 87.5%. In the
Significantly Different With Increased Intensities training and test groups, the positive likelihood ratios
were 9.35 and 6.66, the negative likelihood ratios were 0
MG Group Control Group
M/Z (Mean SD) (Mean SD) p Valuea and 0.191, positive predictive values (PPV) were 1/1071
and 1/1501, and negative predictive values (NPV) were
1122.57 1.756 2.198 0.477 0.431 0.001 100% and 99.99%, respectively.
1282.53 1.536 1.736 0.723 0.436 0.039
1436.04 3.716 3.169 2.271 2.216 0.037
1698.90 1.329 0.874 0.842 0.820 0.008 Comment
2768.64 1.693 1.203 0.715 0.413 0.001 Acquired MG is a rare disorder with 200 to 400 cases/
2778.32 1.577 1.367 0.819 0.409 0.042 million as reported [7]. This prevalence is almost similar
2817.05 2.561 1.937 0.941 0.575 0.001 in China and Western countries [8]. Several approaches
2951.56 4.293 3.810 1.144 0.605 0.001 have been taken to facilitate accurate and early diagnosis
3189.62 4.895 5.094 0.649 0.498 0.001 of MG. Pharmacologic testing with intravenous edropho-
3240.11 1.199 1.041 0.500 0.549 0.004 nium is plagued with the problem of both false-negative
3266.04 5.749 6.755 1.201 1.563 0.001 and false-positive results. The repetitive nerve stimula-
3372.34 2.816 2.065 1.589 0.944 0.018 tion test is relatively insensitive in ocular and in mild
3393.46 8.771 8.114 3.884 1.314 0.015 generalized MG. In addition, nearly half of patients with
3442.53 2.981 2.191 1.460 0.854 0.002 ocular MG are seronegative by acetylcholine receptor
3467.51 0.958 0.861 0.208 0.167 0.001 (AchR) or muscle-specific tyrosine kinase (MuSK) anti-
3484.71 1.436 1.217 0.805 0.604 0.029
bodies tests. Given the shortcomings and disadvantages
of these approaches, there is a dire need for a new
4958.12 5.710 5.356 2.850 1.997 0.042
diagnostic method to diagnose MG early and accurately
5329.16 9.302 8.444 4.284 2.942 0.021
because early and accurate diagnosis is important to
5625.64 18.670 14.127 8.135 4.422 0.002
obtain the best clinical benefit from thymectomy during
5701.79 0.698 0.594 0.278 0.148 0.001
the early course of MG [9, 10].
5895.70 25.846 18.598 7.834 5.568 0.001
In this study, we analyzed the protein spectra in MG
6104.09 2.778 2.595 0.764 0.476 0.001 patients by SELDI TOFMS. Our results suggest that this
6619.48 27.453 16.580 17.658 12.701 0.022 technology may offer a useful tool for the diagnosis of
7959.83 4.293 4.130 1.612 0.897 0.003 MG using serum samples and holds promise to identify
10248.9 0.576 0.445 0.293 0.176 0.008 novel biomarkers of MG that can provide insights in the
15088.6 0.302 0.275 0.143 0.130 0.007 pathophysiology of MG.
15831.7 0.455 0.406 0.183 0.137 0.001 SELDI TOFMS has several potential advantages as a
15911.7 2.161 2.431 0.630 0.514 0.003 clinical assay. This technology provides reliable repro-
16080.2 0.738 0.920 0.199 0.152 0.001 ducible data from readily accessible serum samples, has
16128.0 0.702 0.789 0.199 0.152 0.001 good throughput, poses minimal risk to patients, and is
relatively inexpensive [1113]. Moreover, this technology
a
Values of p 0.05 were significant.
has been successfully used in the diagnosis of many
MG myasthenia gravis; M/Z mass-to-charge ratio; SD diseases by protein profiling of complex biologic mix-
standard deviation.
tures. Because protein chip arrays can selectively bind to

(Fig 2). This tree had the overall lowest relative cross-
validated cost of 0.080, and the cost was considerably less Table 2. Comparison of Protein Peaks That Were
than that of the next smallest tree. The optimal decision Significantly Different With Decreased Intensities
tree used the peak at M4168.94 Da as a primary splitter in
the classification process (Fig 2). Node 1 was split at MG Group Control Group
M/Z (Mean SD) (Mean SD) p Valuea
M4168.94, and a case went left if M4168.94 was 7.233 or
less: 50 cases went left as terminal node 1 and 62 cases 3154.77 1.987 2.236 7.303 4.309 0.001
right as node 2. Node 2 was split at M1122.57, and a case 3951.23 1.839 2.661 6.316 3.421 0.001
went left if M1122.57 was 1.368 or less: 54 cases were 3969.56 0.981 1.339 4.549 3.453 0.001
divided to the left side as terminal node 2 and 8 cases to 4168.94 4.105 4.050 21.163 9.721 0.001
the right side as node 3. In the training data set of 112 4276.40 1.740 2.513 13.431 7.363 0.001
samples (MG and control), the sensitivity and specificity 4468.40 1.850 1.276 2.692 1.505 0.015
of this algorithm were 100% and 89.3%, respectively. 8128.51 1.748 1.508 4.089 3.388 0.002
Furthermore, using a blinded validation set of sera 8596.61 0.963 0.954 2.085 1.335 0.001
samples (test set, 24 MG patients and 24 healthy con- 8921.70 1.410 1.675 3.903 3.325 0.001
trols), the cross-validation analysis showed that 20 MG
patients and 21 healthy controls were correctly identified a
Values of p 0.05 were significant.
following the decision classification tree for an 85.4% MG myasthenia gravis; M/Z mass-to-charge ratio; SD
accuracy rate. For the validation set, the algorithm had a standard deviation.

Downloaded from ats.ctsnetjournals.org by on February 10, 2012

Ann Thorac Surg
2009;88:1118 23
proteins based on the physical or chemical modifications
of their surface, the protein contents of the samples are
thus effectively simplified, and the contaminants such as
buffer salts or detergents can also be easily washed away
before analysis. Compared with 2-dimensional protein
gel electrophoresis (2D gel) and MS, SELDI technology is
much faster with a high throughput capability and is
advantageous in its ability to effectively resolve low-mass
proteins (2000 to 20,000 Da), hydrophobic proteins, and
very basic or acidic proteins [14]. Not only are the
mass/charge values of proteins available but also some
physical and chemical properties of proteins, including
hydrophobicity/hydrophilicity, phosphorylation and
metal-binding.
A particular advantage is that this technique needs
only a very small amount of sample to analyze biologic
mixtures directly or after simple pretreatment. Further-
more, the cross-validation analysis in the training and
test set performed by the Biomarker Pattern Software can
Downloaded from ats.ctsnetjournals.org by on February 10, 2012
CHENG ET AL 1121
GENERAL THORACIC
DIAGNOSING MG BY SERUM PROTEINS
Fig 1. Representative surface-enhanced laser
desorption/ionization spectra comparison of
serum between the myasthenia gravis (MG)
patients and healthy control groups. In the
MG sera, the proteins at M3951.23,
M3969.56, M4168.94, and M4276.40 were
decreased in comparison with the normal con-
trol sera. The x axis shows the molecular
weight for each spectrum (mass/charge ratio
values), and the y axis shows the relative
intensity.
enhance the validation of the performance of the decision
tree and in particular, to evaluate the sensitivity and the
specificity.
Because 41 of the 80 patients in this study were in the
ocular MG class, our data suggest that SELDI-based
serum analysis may be beneficial to the diagnosis for this
subset of patients. In addition, Chinese patients with
generalized MG have a much lower rate of 32% to 63%
[15] of positive serum AchR antibodies compared with
80% to 90% in white patients [16]. This implies that
SELDI-based serum analysis may be of importance for
Chinese patients with generalized MG.
With SELDI TOFMS, we identified 39 protein peaks,
which showed a statistical difference between the MG
group and the healthy controls. Only 2 peaks (M4168.94
and M1122.57 Da) were used in the classification tree
algorithms. The study of Wadsworth and colleagues [17]
suggested that the classification and regression tree ap-
proach examined all the possible protein peak combina-
 1122 CHENG ET AL
GENERAL THORACIC
DIAGNOSING MG BY SERUM PROTEINS
Fig 2. In the training data set, the Biomarker Pattern Software built
the classification tree to classify myasthenia gravis patients or
healthy controls. All the sera were categorized for a sensitivity and
specificity of 100% and 89.3%, respectively, using peak at M4168.94
Da, M1122.57 Da as splitters.
tions from the input spectral data and made the best
classification tree. The classification tree in our study has
shown that the 112 samples in the training set were
classified as either the MG or control group with only two
nodes, with an encouraging outcome of 100% sensitivity
and 89.4% specificity. In the subsequent cross validation
analysis, the sensitivity was 83.3% and specificity was
87.5%.
The high sensitivity and specificity obtained by the
serum protein profiling approach presented in this study
demonstrate that SELDI protein chip MS can be an
intelligent classification algorithm to facilitate the discov-
ery of serum biomarkers for MG and provide an innova-
tive clinical diagnostic platform that has the potential for
early detection and differential diagnosis of MG.
In the 4 nonidentified patients in the MG group, 2
patients had accompanying hyperthyroidism. Whether
hyperthyroidism affects protein peaks is not clear at the
present time. However, no major clinical and biologic
differences were found in the 3 nonidentified individuals
in the control group.
A limitation of this study is that the peaks generated
have not been identified. In the decision tree, the protein
peaks used as the splitters were identified with an
average mass of M4168.94 and M1122.57 Da. MG is an
antibody-mediated autoimmune disease, and some spe-
cial serum antibodies, including AChRs and MuSK, are
often detected in generalized MG [18, 19]. Recently, Leite
and colleagues [20] reported immunoglobulin G1 anti-
bodies to AChRs in seronegative MG patients whose test
results for AChR and MuSK antibodies were negative.
Therefore, the identities of the proteins at M4168.94 and
M1122.57 Da remains to be further clarified.
Another limitation is the lack of gold standard for the
diagnosis of MG. To minimize diagnostic mistakes, we
strictly set the inclusion criteria for the MG patients in
this study to meet the 5 diagnostic criteria mentioned
above. The detection of AChR and MuSK antibodies was
Downloaded from ats.ctsnetjournals.org by on February 10, 2012
Ann Thorac Surg
2009;88:1118 23
not standard routine clinical practice in China when this
study was conducted, but the serology status of the MG
patients will be included in future studies.
Our results suggested that certain proteins might po-
tentially be identified to differentiate MG disease by
SELDI TOFMS analysis. However, a large prospective
clinical trial is still needed to validate the reliability and
diagnostic performance of the decision tree model. Be-
cause MG is markedly influenced by genetic factors [21],
validation in Western countries is also needed.
Furthermore, extended transsternal thymectomies for
MG can obtain complete remission in less than 50% of
patients after long-term follow-up [22]. That means many
patients will not benefit from thymectomy. Therefore, it
is important to use this technology to identify the sub-
population of MG patients who would not benefit from
thymectomy to avoid unnecessary operations. Most of
the MG patients in our study are being monitored to see
if they derive benefit from thymectomy. When the fol-
low-up data are available, the stored sera of these pa-
tients will be analyzed to identify the serum markers that
can predict the therapeutic response to thymectomy.
In conclusion, our results demonstrate that serum
protein profiling using the new technology of SELDI
TOFMS can discriminate patients with MG from healthy
individuals. There is a potential of combining the serum
protein profiles and clinical features for the early diag-
nosis of MG with a high sensitivity and specificity.
We thank Dr Yilong Wu (Lung Cancer Research Institute of
Guangdong Provincial peoples Hospital) for providing the
SELDI TOFMS service. This study was supported by grants from
the National Natural Science Fund of China (30801136 to C.C.,
90713036 to J. Pan), National High Technology Research and
Development Program of China (863 Program 2008AA02Z420 to
J. Pan), and National Basic Research Program of China (973
Program 2009CB825506 to J. Pan).
References
1. Juel VC, Massey JM. Myasthenia gravis. Orphanet J Rare Dis
2007;2:44.
2. Jaretzki A 3rd, Barohn RJ, Ernstoff RM, et al. Myasthenia
gravis: recommendations for clinical research standards.
Task Force of the Medical Scientific Advisory Board of the
Myasthenia Gravis Foundation of America. Neurology 2000;
55:16 23.
3. Han KQ, Huang G, Gao CF, et al. Identification of lung
cancer patients by serum protein profiling using surface-
enhanced laser desorption/ionization time-of-flight mass
spectrometry. Am J Clin Oncol 2008;31:1339.
4. Jansen C, Hebeda KM, Linkels M, et al. Protein profiling of
B-cell lymphomas using tissue biopsies: a potential tool for
small samples in pathology. Cell Oncol 2008;30:2738.
5. Cazares LH, Diaz JI, Drake RR, Semmes OJ. MALDI/SELDI
protein profiling of serum for the identification of cancer
biomarkers. Methods Mol Biol 2008;428:125 40.
6. Osserman KE, Genkins G. Studies in myasthenia gravis:
review of a twenty-year experience in over 1200 patients. Mt
Sinai J Med 1971;38:497537.
7. Robertson DN. Enumerating neurology. Brain 2000;123:
663 4.
8. Yu YL, Hawkins BR, Ip MS, Wong V, Woo E. Myasthenia
gravis in Hong Kong Chinese. 1. Epidemiology and adult
disease. Acta Neurol Scand 1992;86:1139.
Ann Thorac Surg
2009;88:1118 23
9. Monden Y, Nakahara K, Kagotani K, et al. Effects of preop-
erative duration of symptoms on patients with myasthenia
gravis. Ann Thorac Surg 1984;38:28791.
10. Kim HK, Park MS, Choi YS, et al. Neurologic outcomes of
thymectomy in myasthenia gravis: comparative analysis of
the effect of thymoma. J Thorac Cardiovasc Surg 2007;134:
6017.
11. Xiao Z, Adam BL, Cazares LH, et al. Quantitation of serum
prostate-specific membrane antigen by a novel protein bio-
chip immunoassay discriminates benign from malignant
prostate disease. Cancer Res 2001;61:6029 33.
12. Cazares LH, Adam BL, Ward MD, et al. Normal, benign,
preneoplastic, and malignant prostate cells have distinct
protein expression profiles resolved by surface enhanced
laser desorption/ionization mass spectrometry. Clin Cancer
Res 2002;8:254152.
13. Petricoin EF, Ardekani AM, Hitt BA, et al. Use of proteomic
patterns in serum to identify ovarian cancer. Lancet 2002;
359:5727.
14. Bichsel VE, Liotta LA, Petricoin EF 3rd. Cancer proteomics:
from biomarker discovery to signal pathway profiling. Can-
cer J 2001;7:69 78.
INVITED COMMENTARY
In the present article, Cheng and associates [1] seek to
address the interesting and currently debated question of
the accurate diagnostic tests for myasthenia gravis (MG).
Indeed, it contributes to a better understanding of the
close relation between an accurate early diagnosis and
optimum clinical benefit from thymectomy during the
course of MG.
In an elegant cohort comparative study, the authors
sought to analyze the protein spectra in MG patients.
They performed a proteome-wide search for potential
serum protein diagnostic markers for MG using surface-
enhanced laser desorption/ionization time-of-flight
(SELDI-TOF) techniques. To do this, they analyzed sera
samples from 80 MG patients and 80 healthy individuals
(controls). Two-thirds of these (56 MG patients and 56
controls) were used as a training set and analyzed to set
up a decision tree, while samples from the remaining
one-third of the population (24 MG patients and 24
controls) were used as a learning set for cross-validation.
The model was then evaluated by means of a bootstrap
analysis. The results showed that the SELDI-TOF tech-
nique generated 101 proteins peaks in the cohort, and 39
of the peaks were statistically different between the MG
and control groups. In the training set, the decision tree
used 2 peaks as splitters, with a sensitivity of 100% and
specificity of 89.3%. In the learning set, the sensitivity was
83.3% and specificity was 87.5%. The authors concluded
that the SELDI-TOF technique useful to detect and
identify potential serum biomarkers that can diagnose
MG with high sensitivity and specificity.
One might ask why thoracic surgeons should be con-
cerned by the findings of this rigorous study. Not only
does it represent a very large series of patients, diag-
nosed and treated the same way over a short lapse of
time, it also pinpoints a crucial issue concerning medical
care: the identification of a subpopulation of patients who
2009 by The Society of Thoracic Surgeons
Published by Elsevier Inc
Downloaded from ats.ctsnetjournals.org by on February 10, 2012
CHENG ET AL 1123
GENERAL THORACIC
DIAGNOSING MG BY SERUM PROTEINS
15. Cui LY, Guan YZ, Wang H, Tang XF. Single fiber electro-
myography in the diagnosis of ocular myasthenia gravis:
report of 90 cases. Chin Med J (Engl) 2004;117:848 51.
16. Vincent A, McConville J, Farrugia ME, Newsom-Davis J.
Seronegative myasthenia gravis. Semin Neurol 2004;24:125
33.
17. Wadsworth JT, Somers KD, Stack BC Jr, et al. Identification
of patients with head and neck cancer using serum protein
profiles. Arch Otolaryngol Head Neck Surg 2004;130:98 104.
18. Vincent A. Unravelling the pathogenesis of myasthenia
gravis. Nat Rev Immunol 2002;2:797 804.
19. Vincent A, Leite MI. Neuromuscular junction autoimmune
disease: muscle specific kinase antibodies and treatments for
myasthenia gravis. Curr Opin Neurol 2005;18:519 25.
20. Leite MI, Jacob S, Viegas S, et al. IgG1 antibodies to acetyl-
choline receptors in seronegative myasthenia gravis. Brain
2008;131:1940 52.
21. Vandiedonck C, Giraud M, Garchon HJ. Genetics of auto-
immune myasthenia gravis: the multifaceted contribution of
the HLA complex. J Autoimmun 2005;25(suppl):6 11.
22. Shrager JB, Nathan D, Brinster CJ, et al. Outcomes after 151
extended transcervical thymectomies for myasthenia gravis.
Ann Thorac Surg 2006;82:18639.
would not benefit from a given surgery, for whom sur-
gery would thus be unnecessary. For instance, thymec-
tomy is commonly recommended for all patients with
generalized MG [2], whatever the rate of complete remis-
sion (frequently reported less than 50%) after long-term
follow-up.
Cheng and colleagues [1] suggest another pathway for
exploring MG management, and in the burgeoning field
of serum biomarkers, has successfully opened one door
leading to ten more. Future works that use SELDI-TOF
technology and assess follow-up may be the clue to
altering the established decision of therapeutic indica-
tions and, thus, to performing thymectomy only if it
appears to be necessary.
For now, Cheng and associates [1] are to be congratu-
lated on their innovative contribution in this area. Their
results will certainly prove to be most beneficial to the
thoracic surgery community.
Pierre-Emmanuel Falcoz, MD, PhD
Department of Thoracic Surgery
Strasbourg University Hospital
1 Place de lHpital
Nouvel Hpital Civil
Strasbourg, 670091 France
e-mail: pierre-emmanuel.falcoz@wanadoo.fr
References
1. Cheng C, Wu G, Yeung S-CJ, et al. Serum protein profiles in
myasthenia gravis. Ann Thorac Surg 2009;88:1118 23.
2. Backmann K, Burkhardt D, Schreiter I, et al. Thymectomy is
more effective than conservative management for myasthenia
gravis regarding outcome, and clinical improvement. Surgery
2009;145:392 8.
0003-4975/09/$36.00
doi:10.1016/j.athoracsur.2009.06.015
 Serum Protein Profiles in Myasthenia Gravis
Chao Cheng, Guoyong Wu, Sai-Ching J. Yeung, Rong Li, Amos Ela Bella, Jinzhuo
Pang, Fo-tian Zhong, Honghe Luo, Yanli Jin and Jingxuan Pan
Ann Thorac Surg 2009;88:1118-1123
DOI: 10.1016/j.athoracsur.2009.05.032

Updated Information including high-resolution figures, can be found at:

& Services http://ats.ctsnetjournals.org/cgi/content/full/88/4/1118
References This article cites 22 articles, 9 of which you can access for free at:
http://ats.ctsnetjournals.org/cgi/content/full/88/4/1118#BIBL
Citations This article has been cited by 2 HighWire-hosted articles:
http://ats.ctsnetjournals.org/cgi/content/full/88/4/1118#otherarticle
s
Subspecialty Collections This article, along with others on similar topics, appears in the
following collection(s):
Mediastinum
http://ats.ctsnetjournals.org/cgi/collection/mediastinum
Permissions & Licensing Requests about reproducing this article in parts (figures, tables) or
in its entirety should be submitted to:
http://www.us.elsevierhealth.com/Licensing/permissions.jsp or
email: healthpermissions@elsevier.com.
Reprints For information about ordering reprints, please email:
reprints@elsevier.com

Downloaded from ats.ctsnetjournals.org by on February 10, 2012