RESEARCH ARTICLE Discovery and Validation of Biomarkers that … · 2015/6/27 · 81 of pancreatic cysts in clinical practice, the diagnostic accuracy of these techniques is not

1

RESEARCH ARTICLE 1

2

Discovery and Validation of Biomarkers that distinguish mucinous 3

and nonmucinous pancreatic cysts 4

5

Jisook Park1*, Hwan Sic Yun2*, Kwang Hyuck Lee2, Kyu Taek Lee2, Jong Kyun Lee2, and 6

Soo-Youn Lee3,4 7

8

1Samsung Biomedical Research Institute, Samsung Medical Center, Sungkyunkwan 9

University School of Medicine, Seoul, Korea.2Departments of Medicine, Samsung Medical 10

Center, Sungkyunkwan University School of Medicine, Seoul, Korea. 3Department of 11

Clinical Pharmacology and Therapeutics, Samsung Medical Center, Seoul, Korea. 12

4Department of Laboratory & Genetics, Samsung Medical Center, Sungkyunkwan University 13

School of Medicine, Seoul, Korea. 14

15

*These authors contributed equally to this work. 16

17

Running Title: Biomarker for Differential Diagnosis of Pancreatic Cysts 18

19

Keywords: Biomarker, MRM, Pancreatic cyst, Mass spectrometry, Validation 20

21

Corresponding Authors: Jong Kyun Lee, Department of Medicine, Samsung Medical 22

Center, Sungkyunkwan University School of Medicine, 81 Irwon-ro, Gangnam-gu, Seoul 23

135-710, Korea. Phone: +82-2-3410-3407; Fax: +82-2-3410-6983; E-mail: 24

on March 6, 2021. © 2015 American Association for Cancer Research. cancerres.aacrjournals.org Downloaded from

Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited. Author Manuscript Published OnlineFirst on June 29, 2015; DOI: 10.1158/0008-5472.CAN-14-2896

http://cancerres.aacrjournals.org/

2

[email protected]; and Soo-Youn Lee, Department of Laboratory Medicine & 25

Genetics, Samsung Medical Center, Sungkyunkwan University School of Medicine, 81 26

Irwon-ro, Gangnam-gu, Seoul 135-710, Korea. Phone: +82-2-3410-1834; Fax: +82-2-3410-27

2719; E-mail: [email protected] 28

29

Disclosure of Potential Conflicts of Interest 30

The authors declare no conflicts of interest. 31




3

ABBREVIATIONS 32

GeLC-SID-MRM-MS, GeLC-Stable Isotope Dilution-Multiple Reaction Monitoring-mass 33

spectrometry; PIGR, polymeric immunoglobulin receptor; LCN2, lipocalin 2; FCGBP, Fc 34

fragment of IgG binding protein; REG1A, lithostathine-1-alpha; AFM, afamin; CTRC, 35

chymotrypsin C (caldecrin); AMY2B, amylase, alpha 2B (pancreatic); LGALS3BP, lectin, 36

galactoside-binding, soluble, 3 binding protein; CELA3A, chymotrypsin-like elastase family, 37

member 3A; EUS-FNA, endoscopic ultrasound-guided fine needle aspiration; CEA, 38

carcinoembryogenic antigen; DEPs, differentially expressed proteins; IHC, 39

immunohistochemistry; ELISA, enzyme-linked immunosorbent assays; EPI, Enhanced 40

product ion; DP, declustering potential; ROC, receiver operating characteristic; AUC, area 41

under the ROC curve; IPMN, intraductal papillary mucinous neoplasm 42




4

Abstract 43

The use of advanced imaging technologies for the identification of pancreatic cysts has 44

become widespread. However, accurate differential diagnosis between mucinous cysts (MC) 45

and nonmucinous cysts (NMC) consisting of pseudocysts (NMC1) and nonmucinous 46

neoplastic cysts (NMC2) remains a challenge. Thus, it is necessary to develop novel 47

biomarkers for the differential diagnosis of pancreatic cysts. An integrated proteomics 48

approach yielded differentially expressed proteins in MC that were verified subsequently in 49

99 pancreatic cysts (21 NMC1, 41 NMC2 and 37 MC), using a method termed GeLC-stable 50

isotope dilution-multiple reaction monitoring-mass spectrometry (GeLC-SID-MRM-MS) 51

along with established immunoassay techniques. We identified 223 proteins by nanoscale 52

liquid chromatography coupled to tandem mass spectrometry (nano LC-MS/MS). Nine 53

candidate biomarkers were identified, including polymeric immunoglobulin receptor (PIGR), 54

lipocalin 2 (LCN2), Fc fragment of IgG binding protein (FCGBP), lithostathine-1-alpha 55

(REG1A), afamin (AFM), chymotrypsin C (caldecrin) (CTRC), amylase, alpha 2B 56

(pancreatic) (AMY2B), lectin, galactoside-binding, soluble, 3 binding protein (LGALS3BP) 57

and chymotrypsin-like elastase family, member 3A (CELA3A), which were established as 58

biomarker candidates for MC. In particular, we showed a biomarker subset including AFM, 59

REG1A, PIGR and LCN2 could differentiate MC not only from NMC (including NMC1) but 60

also from NMC2. Overall, the MS-based comprehensive proteomics approach employed in 61

this study established a novel set of candidate biomarkers that address a gap in efforts to 62

distinguish early pancreatic lesions at a time when more successful therapeutic interventions 63

may be possible. 64

65

66




5

Introduction 67

Recently, the use of ultrasound (US), multidetector computed tomography (CT), and 68

magnetic resonance imaging (MRI) techniques have been most commonly used for the 69

detection of pancreatic cysts. The prevalence of pancreatic cysts within the population ranges 70

from 2.6% to as high as 44.7% worldwide (1-4), and a study from Japan reported that 24.3% 71

of the population has pancreatic cysts based on autopsy cases. However, an accurate 72

differential diagnosis of cystic pancreatic lesions remains a challenge. 73

Accurate differential diagnosis of mucinous cysts (MC) from nonmucinous cysts 74

(NMC) including pseudocysts (NMC1) and nonmucinous neoplastic cysts (NMC2) is 75

extremely important, because MC has malignant potential and requires surgical resection. 76

Usually NMC1 are easily differentiated from MC because of their typical history and imaging 77

characteristics, but sometimes these cysts are difficult to differentiate without histology. 78

Although endoscopic ultrasound (EUS), EUS-guided fine needle aspiration (EUS-79

FNA), and cyst fluid analysis all provide additional information for the differential diagnosis 80

of pancreatic cysts in clinical practice, the diagnostic accuracy of these techniques is not 81

satisfactory. The diagnostic accuracy of CT or MRI ranges from 55-75% depending on the 82

reviewers. The accuracy of EUS morphology, cytology, and cystic carcinoembryogenic 83

antigen (CEA) ranges from 50-79% (5). 84

The successful completion of human genome projects has led to the development of 85

various proteomics-based approaches used to identify novel potential biomarkers that are 86

capable of predicting diseases at early stages with minimal invasiveness (6-10). In fact, recent 87

technical advances in mass spectrometry (MS) have been able to detect a large number of 88

low-abundance proteins in biological specimens, which results in the ability to identify many 89

protein surrogates that may indicate malignant properties. As such, this is an attractive 90

method to easily monitor global proteome alterations to identify meaningful protein 91




6

differences among disease states. Identifying proteomic differences in pancreatic cysts may 92

provide differences between malignant tumors and other controls due to its proximity to the 93

disease site (11). 94

Currently, the analysis of cyst fluid CEA is considered the best discriminating tool 95

for mucinous and nonmucionous lesions of the pancreas; however, obtaining sufficient 96

amounts of cyst fluid is not easy for accurate CEA analysis. Furthermore, the utility of cyst 97

fluid CEA has been limited due to its insufficient sensitivity or specificity. On the other hand, 98

a proteomic analysis would require only small amounts of cyst fluid for testing, which would 99

be a distinct advantage. Several proteins have been suggested as potential biomarkers to 100

differentiate potentially malignant MC from NMC (11-15). Previous work on the global 101

proteome of pancreatic cysts using mass spectrometry has discovered novel biomarkers and 102

several biomarker candidates (e.g. olfactomedin-4, mucin-18 and solubilized CEA-related 103

cell adhesion molecules) that were differentially expressed in malignant lesions. To verify 104

these biomarker candidates, immunohistochemistry (IHC) or enzyme-linked immunosorbent 105

assays (ELISA) were performed. However, most of the biomarker candidates identified could 106

not be sufficiently validated as different between patients and healthy donors. Antibody-based 107

approaches have generally been used to verify an identified novel biomarker, but several 108

limitations of these platforms should be carefully considered, including i) the large amount of 109

sample needed in cases where of a large number of required proteins screens, ii) antibody 110

problems in developing a suitable experimental design, and iii) the difficulty of obtaining 111

high quality antibodies. 112

Therefore, in this study, we performed a MS-based approach to overcome these 113

obstacles. We identified potential proteomic marker candidates discriminating MC from other 114

pancreatic cysts by differential proteomics analysis. For biomarker validation, a GeLC-Stable 115

Isotope Dilution-Multiple Reaction Monitoring-mass spectrometry (GeLC-SID-MRM-MS), 116




7

which is a very sensitive and specific proteomics-based quantitative method, allowed for the 117

detection of protein at low attomole concentrations with a minimal amount of specimen 118

consumed without the need for the use of antibodies. 119




8

Methods 120

Study subjects 121

The Institutional Review Board at Samsung Medical Center (Seoul, Korea) approved 122

this study. The patients were recruited prospectively from February 2008 to January 2015. 123

Surgical candidates were directly consulted for surgery. Other patients, who were assumed to 124

have benign cysts with thin walls of small size (less than 2.0 cm) or typical serous cyst 125

adenomas on CT or EUS, underwent follow-up on a regular basis without EUS-FNA. 126

Enrolled patients with pancreatic cysts underwent EUS-FNA. A Hitachi-Aloka® with 127

sector scan transducer (5 MHz) was used for the linear endoscopic ultrasounds. Two 128

specialized endoscopic ultrasonographers performed each EUS-FNA. Aspirated pancreatic 129

cysts were described in terms of location, size, viscosity and fluid volume. Each sample was 130

subjected to cytology, tumor marker identification and chemical analysis. Pancreatic cyst 131

fluids were harvested, frozen immediately and stored at – 70 ℃ for further cytological 132

examination and proteomic analysis. 133

134

Sample preparation for proteomics experiments 135

A MARS column (Agilent Technologies, Santa Clara, CA, USA) was used to 136

remove six abundant proteins (albumin, transferring, IgF, IgA, anti-trypsin and haptoglobin) 137

from pancreatic cyst fluid while all other proteins were enriched. Seventy micrograms of 138

depleted cyst fluid was resolved on 4 - 12% NuPAGE gels to reduce the complexity. Briefly, 139

Coomassie blue-stained protein lanes on the SDS-PAGE gel were manually cut into 18 bands 140

and subjected to in-gel tryptic digestion prior to LC-MS/MS. 141

For GeLC-SID-MRM-MS experiments, cyst fluid was fractionated on SDS-PAGE 142

and stained with Coomassie blue and then cut into 3 bands followed by in-gel trypsin 143




9

digestion at a 50:1 ratio for 16 h at 37℃. The tryptic digests were recovered by extraction 144

with 50% acetonitrile (ACN) / 0.1% formic acid and were dried. Synthetic peptides 145

containing stable-isotope labels were spiked in tryptic digests of cyst fluid as standards (100 146

Fmole/uL of each peptide). These mixtures were purified using an OMIX C18 tip (Agilent, 147

Technologies, Santa Clara, CA, USA). Samples were dried and reconstituted with 10 uL 0.1% 148

TFA in water. 149

150

Mass spectrometry-based experiments 151

For the proteomic profiling of pancreatic cyst fluids, MS analysis of cyst fluid tryptic 152

digests were performed using a 6520 quadrupole time-of-flight (Q-TOF) mass spectrometer 153

(Agilent Technologies, Santa Clara, CA, United States). The Q-TOF LC/MS was operated at 154

high resolution (4 GHz) on positive ion mode for all experiments. All data was acquired in 155

the data-dependent MS/MS mode, where a full MS scan was followed by three MS/MS scans 156

for the three most abundant precursor ions in the MS survey scan. 157

For the determination of the cystic proteome of interest, we achieved MRM using a 158

QTRAP 5500 (ABSciex, Framingham, MA, USA) equipped with a nano-electrospray ion 159

source. A MRM scan was performed on the positive mode with ion spray voltages in a 160

2,200~2,500 V range. 161

162

Proteomic data acquisition 163

All MS/MS spectra were searched against the human Swiss-Prot database using the 164

Spectrum Mill search engine (rev A.03.03.084 SR4, Agilent Technologies, Santa Clara, CA, 165

USA). Search results were automatically validated using Spectrum Mill software (Agilent 166

Technologies, Santa Clara, CA, USA) to meet a false discovery rate (FDR) < 1%. Peptides 167

with a 1% FDR cutoff were used to identify proteins via automatic validation using the 168




10

following parameters: scored percent intensity (SPI) > 70% for matches with scores > 13 for 169

+1, > 11 for +2, and > 13 for +3. Then, proteins were filtered based on the following criteria: 170

SPI > 70% for matches with scores > 6 for 1+, 2+ and 3+, rank1-2 score threshold = 2, and 171

protein score > 20. Label free and relative semi-quantification by spectral counts, which are 172

useful for quantifying protein abundance changes in shotgun proteomics, were performed for 173

biomarker discovery. The ratio of spectral counts (Rsc-value) (16), which is the log2 (protein 174

ratio) measured from spectral counts, was used for statistical validation of the proteomic data 175

acquired. The Rsc > 1 indicates a 2-fold change in group comparison was allowed for the first 176

screen of the protein list. 177

MRM transitions generated by MRMPilot v. 2.0 (AB SCIEX, Framingham, MA, 178

USA) used were monitored for MRM experiments. At least three transitions from one 179

proteotrypic peptide were generated; 2 or 3 peptide charge states containing 8-25 amino acids, 180

no post-translational modification (PTM), and one missed cleavage were allowed. 181

182

Statistical analysis 183

Data acquired from targeted quantitative proteomics were analyzed using the SPSS 184

statistical package (IBM Corporation, Somers, NY, USA), ver 21.0 and MedCalc software ver 185

12.7.0.0 (Mariakerke, Belgium). When the distribution of data was not normal (i.e., 186

proteomic markers), non-parametric tests were used and data with a normal distribution (i.e., 187

gender and age) were analyzed with parametric tests. Differences in the concentration of 188

proteomic markers between groups were assessed with the non-parametric Mann–Whitney 189

test. The Chi-square (χ2) test was used to assess differences in gender between groups. A T-190

test was conducted to assess age between groups. A P-value ≤ 0.05 was considered 191

statistically significant. ROC analysis was used to access the diagnostic performance of 192

proteomic markers and the area under the curve (AUC) was estimated. 193




11

Results 194

Subject characteristics 195

Ninety-two pancreatic cysts were collected prospectively between 2008 and 2012 at 196

Samsung Medical Center. A shortage of cystic fluid (n = 15), not belonging to one of the 197

subgroups (n = 2) or indeterminate diagnosis (n = 21) were the reasons for exclusion from the 198

study. Finally, a total of 99 pancreatic cysts were subjected to analysis and classified into 199

three subgroups (21 NMC1, 41 NMC2 and 37 MC; 38 NMC and 18 MC were included in the 200

validation cohort 1 and 24 NMC and 19 MC in the validation cohort 2) (Table 1). The mean 201

age of the patients was 52 years (range, 20-76 years). No statistically significant differences 202

were observed in age or sex among the subgroups. The amount of aspiration fluid ranged 203

from 1 ㎖ to 120 ㎖, with an average fluid volume of 19.9 ㎖. The most common location 204

of the cysts among the patients was in the pancreatic tail and head (64.3%). Fourteen (25.0%) 205

and six (10.7%) cysts were located in the pancreatic body and peri-pancreas, respectively. 206

Only one patient experienced pancreatitis after EUS-FNA. 207

The final diagnosis was made based on histological and cytological findings. For 208

patients in whom tissue could not be obtained, CT, MRI, EUS findings, cystic fluid CEA, 209

clinical history and cystic change during the follow-up period were used to make a final 210

clinical diagnosis (Supplementary Fig. S1) (17). MC were diagnosed on histology or based 211

on any 2 of the following characteristics: 1) cystic fluid CEA > 192 ng/mL (5), 2) EUS 212

consistent with intraductal papillary mucinous neoplasm (IPMN) or MCN, 3) cytology 213

consistent with MC. NMC2 were diagnosed on histology or based on any 2 of the following 214

characteristics: 1) cystic fluid CEA <5 ng/mL, 2) EUS consistent with NMC, 3) cytology 215

consistent with NMC. Pseudocysts were diagnosed on histology based on EUS consistent 216

with pseudocyst and a typical clinical history including cyst resolution on follow-up. 217




12

CEAs were measured using a radioimmunoassay kit (RIAKEY®, Shin Jin Medics 218

Inc. Seoul, Korea). Histological or cytological diagnoses were made in 44 patients, and 219

clinical diagnoses were made in 56 patients. 220

221

Biomarker identification 222

The analytical strategies of biomarker development in pancreatic cysts are shown in 223

Fig. 1. We identified 223 proteins (Supplementary Table S1) and performed semi-224

quantification by spectral counts, which represent the abundance of each protein, with the 225

aim to identify differentially expressed proteins (DEPs) (Supplementary Table S2). In the 226

current study, we paid attention to a set of proteins that changed in MC with high malignant 227

potential. Thus, we chose proteins with Rsc > 1 value that indicates a 2-fold change in each 228

group compared to the malignant group. Finally, we identified 62 DEPs among the disease 229

groups (NMC1, NMC2, MC and malignancy). 230

231

Biomarker prioritization 232

We used MRM assay to prioritize a set of proteins of interest derived from the 233

biomarker identification using shotgun proteomics and select specific peptides from these 234

markers suitable for the MRM assay. A total of 1,021 MRM transitions against 62 DEPs were 235

created and then MRM experiments were achieved on target peptides screened through the 236

first discovery process using QTRAP 5500 MS. Based on this analysis, 33 proteins 237

consequently satisfied the filtering criteria on pooled pancreatic cyst samples. Briefly, a 238

reproducible chromatographic retention time property and strong MS signal intensity were 239

considered as follows: high MS response (signal to noise ratio > 8) and high reproducibility 240

(± 20 s retention time tolerance of three transitions). A total of 33 protein levels were 241

relatively quantified on 56 pancreatic cysts (17 NMC1, 21 NMC2 and 18 MC) using MRM-242




13

MS without IS for initial screening. A disease specificity and diagnostic performance of 243

candidates were referred for the second triage, and specific biomarkers were selected using 244

Kruskal-Wallis tests and ROC analysis. After this assessment, 17 proteins were found to be 245

significantly altered in MC compared to NMC1 or NMC2 (P < 0.05 and AUC > 0.7). 246

Several proteins found in the discovery stage of our study overlapped with a set of 247

biomarker surrogates previously reported in pancreatic cysts (Supplementary Table S3). 248

However, these proteins were not specific for mucinous cysts and were not measureable 249

using MRM assay; thus, they were eliminated in a further validation phase. 250

251

Biomarker validation 252

Of the 17 proteins isolated, nine proteins (LGALS3BP, PIGR, LCN2, FCGBP, 253

REG1A, AFM, CTRC, AMY2B and CELA3A) were finally selected for the validation phase 254

based on the preliminary results by MRM experiments with isotope labeled peptides. Herein, 255

we conducted SDS-PAGE prior to MRM experiments in order to reduce the complexity of 256

cyst fluids and to increase the sensitivity for the target proteins of interest (named GeLC-SID-257

MRM-MS). Their concentrations were then measured (Table 2, Fig. 2) in cyst fluids using 258

GeLC-SID-MRM-MS. In the validation cohort 1, the concentration of the 7 proteins (PIGR, 259

LCN2, FCGBP, REG1A, CTRC, AMY2B and CELA3A) was higher in MC than NMC2 by 260

approximately 7-fold. AFM levels were significantly decreased by 6-8-fold in MC (41 261

amol/㎕) compared to NMC1 (260.7 amol/㎕) or NMC2 (340 amol/㎕), respectively. 262

Specifically, four proteins, AFM, FCGBP, REG1A and AMY2B, were found to be 263

significantly altered in MC compared to NMC (including NMC1 and NMC2; P < 0.05) 264

(Table 3, Fig. 2). Thus, we additionally validated 4 proteins in a separate validation cohort 2. 265

These markers also demonstrated surprisingly good diagnostic performance for 266




14

discriminating MC and NMC (P < 0.001) as follows: AUCAFM=0.839 (95% CI= 0.687 to 267

0.937), AUCLCN2=0.895 (95% CI= 0.754 to 0.970), AUCREG1A=0.945 (95% CI= 0.821 to 268

0.992) and AUCPIGR=0.887 (95% CI= 0.744 to 0.966). Here, a biomarker subset including 269

AFM, FCGBP, REG1A and AMY2B showed good discriminatory performance in not only 270

the validation cohort 1, but also the validation cohort 2. 271

In addition to the results of SID-MRM-MS, we performed immunoassays to detect 272

LCN2, which had previously been shown to be present at higher levels in MC. Fig. 3 shows 273

that LCN2 levels are increased in MC by both methods. The Spearman correlation coefficient 274

was used to determine a correlation between SID-MRM-MS and immunoassays (r = 0.499, P 275

= 0.0014). 276

277

Evaluation of multiple biomarkers 278

A total of 9 proteins were finally selected that showed differential expression in 279

cystic fluids relative to NMC1, NMC2 and MC. To enhance the diagnostic power, we 280

assessed the diagnostic performance, such as sensitivity, specificity, accuracy and AUC 281

among the three groups. These proteomic markers had a wide range of approximately 44~100% 282

and 47~95% for sensitivity and specificity, respectively (Supplementary Table S4, Table 3). 283

Of these, four markers (AFM, REG1A, LCN2 and PIGR) selected through the validation 284

cohort 1 also significantly differentiated MC from NMC (P < 0.001, AUC > 0.8) in the 285

validation cohort 2. Interestingly, PIGR and LCN2 were specific to discrimination of MC 286

from NMC and NMC2, respectively. Thus, we generated biomarker panel 1(AUC, 0.919) by 287

combining AFM, REG1A and LCN2 for NMC2 vs. MC. AFM, REG1A and PIGR were 288

chosen for differentiating MC from NMC as panel 2 (AUC, 0.933) (Fig. 4). 289

To evaluate the diagnostic performance of four markers, we calculated the diagnostic 290

probability score (PS) based on the sum of the hit numbers of 5 proteins, including cystic 291




15

CEA, over cutoff values ranging from 0~5. We described PS with (PS+) and without (PS-) a 292

CEA cutoff in both validation cohorts and the cutoff PS value of 3 was adopted for 293

differential diagnosis of MC. Using a PS- cutoff, 25 of 37 patients with MC were diagnosed 294

with 80% accuracy (Supplementary Table S5). Moreover, when a PS+ cutoff was applied to 295

both validation cohorts, the performance was enhanced, with 30 of 37 MC patients accurately 296

diagnosed, representing 81% sensitivity, 90% specificity and 86% accuracy. We strongly 297

suggest that a 3 ≤ PS is indicative of high MC diagnostic potential. 298




16

Discussion 299

Pancreatic cysts have a wide spectrum of disease states ranging from benign to 300

malignant. Differential diagnosis between NMC and MC is crucial to determining the most 301

appropriate treatment because MC has malignant potential. Cystic fluid CEA, EUS-FNA, CT 302

and MRI have been used to aid in determining the final diagnosis in clinical practice, but 303

their diagnostic accuracies are not satisfactory. Cystic fluid cytology has low sensitivity due 304

to scant cell numbers in samples. CEA, a known clinical tumor marker, is also often 305

suggested as a differential diagnostic marker for MC (5, 18-20). However, the optimal cut-off 306

value of cystic fluid CEA ranges widely, between 109.9 ng/㎖ and 480 ng/㎖ depending on 307

the study. Its diagnostic accuracy for differentiating MC and NMC ranges from 72% to 86% 308

in the literature. Although the primary aim of EUS-FNA is to determine malignancy, the 309

diagnostic yield of EUS-FNA in this study was very disappointing (52.6%). Additionally, CT 310

or MRI images were included in the characterization of cystic pancreatic masses (55-75%). 311

Thus, high quality diagnostic and prognostic tools are required allowing for minimally 312

invasive testing with better accuracy. 313

For decades, various MS techniques have increased for diagnostic use in clinical 314

laboratories for clinical application. Despite these efforts, MS-based approaches are being 315

used more often to identify small molecules rather than protein markers. In fact, only a few 316

proteomic investigations of pancreatic cyst fluids have been applied to overcome the current 317

diagnostic limitations (11, 13, 15) for the differential diagnosis of cysts and early prediction 318

of pancreatic cancer. Nevertheless, no reliable biomarkers have been discovered that can 319

discriminate MC with malignant potential from NMC. Thus, our group focused on 320

identifying proteomic markers that are differentially expressed in MC with high malignant 321

potential compared with NMC, which would be helpful for the differential diagnosis of 322




17

pancreatic cysts. 323

We analyzed pancreatic cyst fluids using integrated proteomics and described the 324

differential expression profiles in different cystic lesions including NMC1, NMC2 and MC. 325

Specially, 9 proteins of interest (PIGR, LCN2, FCGBP, REG1A, AFM, CTRC, AMY2B, 326

LGALS3BP and CELA3A) were verified using MS-based targeted proteomics technologies 327

and immunoassays. These markers have been previously described as potential biomarker 328

candidates associated with multiple cancer types in other studies. For instance, AFM was 329

reported as a potential biomarker for ovarian cancer (21) and breast cancer (22), where AFM 330

may play a role in the progression of cancer. Plasma AFM concentrations were validated by 331

western blotting and immunoassays in patients with ovarian cancer. Serum REG1A 332

concentrations were shown to be significantly over-expressed in patients with various 333

pancreatic diseases, cirrhosis and various cancers of the digestive system (23). Abnormal 334

expression of PIGR was reported in hepatocellular carcinoma (24) and lung cancer (25). 335

LCN2 has been proposed as an early screening marker for human primary breast cancer (26, 336

27), ovarian cancer (28), colorectal cancer and pancreatic cancer (29-31). The exact 337

mechanism for how these proteins correlate with pathogenesis has yet to be clearly 338

understood, despite that these proteins have been suggested as potential biomarkers in various 339

cancers. 340

Our data also showed that 3 proteins (REG1A, PIGR and LCN2) were up-regulated 341

and AFM was down-regulated in MC compared with NMC2. This was in agreement with 342

previous reports that showed alteration of these protein levels depending on malignant 343

potential. In particular, PIGR protein was consistently up-regulated in MC in comparison 344

with not only NMC1 but also NMC2. NMC1 are inflammatory cysts and are not difficult to 345

differentiate from NMC2 because the clinical history and course are different. Some 346

inflammatory markers could be also up-regulated in MC. Therefore, useful markers of MC 347




18

need to be altered discriminately in both NMC1 and NMC2. 348

As mentioned above, cross-sectional imaging (e.g. CT, MRI) or conventional tumor 349

markers (e.g. CEA, CA 19-9, CA 72-4) for the characterization of pancreatic cyst lesions 350

have not provided effective diagnoses in previous studies. In contrast, our data show that the 351

4 marker candidates (AFM, REG1A, PIGR, and LCN2) and the combination of these proteins 352

(panel 1, panel 2) can differentiate MC from NMC groups. Furthermore, the multi-marker 353

panel has greater discriminatory power for NMC versus MC (AUC, 0.933) than that of each 354

single marker (AUC, < 0.829) in the validation phase. 355

Our study demonstrates the usefulness of MS-based comprehensive proteomics 356

methodologies for the discovery and validation of candidate biomarkers for differential 357

diagnosis of pancreatic cysts. Multiple biomarkers (AFM, REG1A, PIGR, LCN2) have been 358

identified, but their reliability and usefulness need to be confirmed in a clinical setting using 359

independent patient cohorts. 360




19

361

Acknowledgments 362

This study was supported by a grant of Korea Health Technology R&D Project 363

through the Korea Health Industry Development Institute (KHIDI), funded by the Ministry of 364

Health & Welfare, Republic of Korea (grant number : HI13C2098) and Samsung Medical 365

Center Clinical Research Development Program grant, #CRS110-56-2.366




20

References

1. Lee KS, Sekhar A, Rofsky NM, Pedrosa I. Prevalence of incidental pancreatic cysts

in the adult population on MR imaging. Am J Gastroenterol 2010;105:2079-84.

2. de Jong K, Nio CY, Hermans JJ, Dijkgraaf MG, Gouma DJ, van Eijck CH, et al.

High prevalence of pancreatic cysts detected by screening magnetic resonance

imaging examinations. Clin Gastroenterol Hepatol 2010;8:806-11.

3. Laffan TA, Horton KM, Klein AP, Berlanstein B, Siegelman SS, Kawamoto S, et al.

Prevalence of unsuspected pancreatic cysts on MDCT. AJR Am J Roentgenol

2008;191:802-7.

4. Girometti R, Intini S, Brondani G, Como G, Londero F, Bresadola F, et al. Incidental

pancreatic cysts on 3D turbo spin echo magnetic resonance

cholangiopancreatography: prevalence and relation with clinical and imaging

features. Abdom Imaging 2011;36:196-205.

5. Brugge WR, Lewandrowski K, Lee-Lewandrowski E, Centeno BA, Szydlo T, Regan

S, et al. Diagnosis of pancreatic cystic neoplasms: a report of the cooperative

pancreatic cyst study. Gastroenterology 2004;126:1330-6.

6. Krieg RC, Paweletz CP, Liotta LA, Petricoin EF, 3rd. Clinical proteomics for cancer

biomarker discovery and therapeutic targeting. Technol Cancer Res Treat

2002;1:263-72.

7. Liang S, Xu Z, Xu X, Zhao X, Huang C, Wei Y. Quantitative proteomics for cancer

biomarker discovery. Comb Chem High Throughput Screen 2012;15:221-31.

8. Liu NQ, Braakman RB, Stingl C, Luider TM, Martens JW, Foekens JA, et al.

Proteomics pipeline for biomarker discovery of laser capture microdissected breast

cancer tissue. J Mammary Gland Biol Neoplasia 2012;17:155-64.




21

9. Lu M, Faull KF, Whitelegge JP, He J, Shen D, Saxton RE, et al. Proteomics and mass

spectrometry for cancer biomarker discovery. Biomark Insights 2007;2:347-60.

10. Haab BB, Porter A, Yue T, Li L, Scheiman J, Anderson MA, et al. Glycosylation

variants of mucins and CEACAMs as candidate biomarkers for the diagnosis of

pancreatic cystic neoplasms. Ann Surg 2010;251:937-45.

11. Ke E, Patel BB, Liu T, Li XM, Haluszka O, Hoffman JP, et al. Proteomic analyses of

pancreatic cyst fluids. Pancreas 2009;38:e33-42.

12. Scarlett CJ, Samra JS, Xue A, Baxter RC, Smith RC. Classification of pancreatic

cystic lesions using SELDI-TOF mass spectrometry. ANZ J Surg 2007;77:648-53.

13. Cuoghi A, Farina A, Z'Graggen K, Dumonceau JM, Tomasi A, Hochstrasser DF, et al.

Role of proteomics to differentiate between benign and potentially malignant

pancreatic cysts. J Proteome Res 2011;10:2664-70.

14. Jabbar KS, Verbeke C, Hyltander AG, Sjovall H, Hansson GC, Sadik R. Proteomic

mucin profiling for the identification of cystic precursors of pancreatic cancer. J Natl

Cancer Inst 2014;106:djt439.

15. Mann BF, Goetz JA, House MG, Schmidt CM, Novotny MV. Glycomic and

proteomic profiling of pancreatic cyst fluids identifies hyperfucosylated lactosamines

on the N-linked glycans of overexpressed glycoproteins. Mol Cell Proteomics

2012;11:M111.015792.

16. Old WM, Meyer-Arendt K, Aveline-Wolf L, Pierce KG, Mendoza A, Sevinsky JR, et

al. Comparison of label-free methods for quantifying human proteins by shotgun

proteomics. Mol Cell Proteomics 2005;4:1487-502.

17. Shen J, Brugge WR, Dimaio CJ, Pitman MB. Molecular analysis of pancreatic cyst

fluid: a comparative analysis with current practice of diagnosis. Cancer

2009;117:217-27.




22

18. Cizginer S, Turner BG, Bilge AR, Karaca C, Pitman MB, Brugge WR. Cyst fluid

carcinoembryonic antigen is an accurate diagnostic marker of pancreatic mucinous

cysts. Pancreas 2011;40:1024-8.

19. Linder JD, Geenen JE, Catalano MF. Cyst fluid analysis obtained by EUS-guided

FNA in the evaluation of discrete cystic neoplasms of the pancreas: a prospective

single-center experience. Gastrointest Endosc 2006;64:697-702.

20. Park WG, Mascarenhas R, Palaez-Luna M, Smyrk TC, O'Kane D, Clain JE, et al.

Diagnostic performance of cyst fluid carcinoembryonic antigen and amylase in

histologically confirmed pancreatic cysts. Pancreas 2011;40:42-5.

21. Jackson D, Craven RA, Hutson RC, Graze I, Lueth P, Tonge RP, et al. Proteomic

profiling identifies afamin as a potential biomarker for ovarian cancer. Clin Cancer

Res 2007;13:7370-9.

22. Opstal-van Winden AW, Krop EJ, Karedal MH, Gast MC, Lindh CH, Jeppsson MC,

et al. Searching for early breast cancer biomarkers by serum protein profiling of pre-

diagnostic serum; a nested case-control study. BMC Cancer 2011;11:381.

23. Shen J, Person MD, Zhu J, Abbruzzese JL, Li D. Protein expression profiles in

pancreatic adenocarcinoma compared with normal pancreatic tissue and tissue

affected by pancreatitis as detected by two-dimensional gel electrophoresis and mass

spectrometry. Cancer Res 2004;64:9018-26.

24. Ai J, Tang Q, Wu Y, Xu Y, Feng T, Zhou R, et al. The role of polymeric

immunoglobulin receptor in inflammation-induced tumor metastasis of human

hepatocellular carcinoma. J Natl Cancer Inst 2011;103:1696-712.

25. Ocak S, Pedchenko TV, Chen H, Harris FT, Qian J, Polosukhin V, et al. Loss of

polymeric immunoglobulin receptor expression is associated with lung

tumourigenesis. Eur Respir J 2012;39:1171-80.




23

26. Bauer M, Eickhoff JC, Gould MN, Mundhenke C, Maass N, Friedl A. Neutrophil

gelatinase-associated lipocalin (NGAL) is a predictor of poor prognosis in human

primary breast cancer. Breast Cancer Res Treat 2008;108:389-97.

27. Fernandez CA, Yan L, Louis G, Yang J, Kutok JL, Moses MA. The matrix

metalloproteinase-9/neutrophil gelatinase-associated lipocalin complex plays a role in

breast tumor growth and is present in the urine of breast cancer patients. Clin Cancer

Res 2005;11:5390-5.

28. Lim R, Ahmed N, Borregaard N, Riley C, Wafai R, Thompson EW, et al. Neutrophil

gelatinase-associated lipocalin (NGAL) an early-screening biomarker for ovarian

cancer: NGAL is associated with epidermal growth factor-induced epithelio-

mesenchymal transition. Int J Cancer 2007;120:2426-34.

29. Brand RE, Nolen BM, Zeh HJ, Allen PJ, Eloubeidi MA, Goldberg M, et al. Serum

biomarker panels for the detection of pancreatic cancer. Clin Cancer Res

2011;17:805-16.

30. Moniaux N, Chakraborty S, Yalniz M, Gonzalez J, Shostrom VK, Standop J, et al.

Early diagnosis of pancreatic cancer: neutrophil gelatinase-associated lipocalin as a

marker of pancreatic intraepithelial neoplasia. Br J Cancer 2008;98:1540-7.

31. Tong Z, Kunnumakkara AB, Wang H, Matsuo Y, Diagaradjane P, Harikumar KB, et al.

Neutrophil gelatinase-associated lipocalin: a novel suppressor of invasion and

angiogenesis in pancreatic cancer. Cancer Res 2008;68:6100-8.




24

Figure legends

Figure 1. Brief strategies for the discovery and validation of candidate biomarkers; In the

discovery phase, label free and relative semi-quantification by spectral counts, which are

useful for quantifying protein abundance changes in shotgun proteomics, were performed for

biomarker discovery (Rsc > 1 was allowed). The MRM-MS without IS was achieved for

initial candidate marker selection. Logistic regression analysis with Bonferroni's correction

and receiver operating characteristic (ROC) curves were used to prioritize identified marker

candidates (P < 0.05, AUC >0.7). The selected candidates were then subjected to the

validation phase (cohort 1 and cohort 2) using GeLC-SID-MRM-MS including IS responding

to 9 proteomic makers. AFM, REG1A, PIGR and LCN2 was finally selected for the

diagnostic panel development.

Figure 2. Concentrations of 9 biomarkers in cystic fluids from patients diagnosed with

NMC1, NMC2 and MC.

Figure 3. Measurement of lipocalin 2 (LCN2) in cystic fluids using (A) immunoassay and (B)

GeLC-Stable Isotope Dilution-Multiple Reaction Monitoring (GeLC-SID-MRM) techniques.

Figure 4. ROC curves for multi-marker panel 1 (A) in NMC2 and MC and panel 2 (B) in

NMC and MC.

Supplementary Fig. S1. Flow chart of study subjects for the validation cohort 1 (A) and cohort




25

2 (B).




26

Table 1. Baseline clinical characteristics 1

Characteristic

Validation cohort 1 Validation cohort 2

NMCa (n = 38) MCd (n = 18)

NMC (n = 24) MC (n = 19)

NMC1b (n = 17) NMC2c (n = 21) NMC1 (n = 4) NMC2 (n = 20)

Age, years Median 57 49 56 44 50 60

Range 20-74 23-67 38-76 25-52 32-65 33-85

Sex, N

Male / Female 11/6 6/15 7/11 3/1 4/16 5/14

EUS findings

Size, mm 5.5 (3-15) 3.2 (2-10) 3.5 (2-12) 6.5 (3-9) 3.5(2-4) 3.7(2-7)

Unilocular/Multilocular 15/2 6/15 4/14 4/0 9/11 10/9

Aspiration, cc 6 (scanty-150) 8.5 (scanty-120) 3 (scanty-40) 12(8-20) 7(1-15) 11 (1-70)

Location of cyst, N (%)

head 2 (11.8) 8 (38.1) 6 (33.3) 0 (0) 10 (50) 6 (31.6)

body and neck 3 (17.6) 8 (38.1) 3 (16.7) 1 (25) 5 (25) 6 (31.6)

tail 8 (47.1) 3 (14.3) 9 (50.0) 3 (75) 5 (25) 7 (36.5)

peri-pancreas 4 (23.5) 2 (9.5) 0 (0) 0 (0) 0 (0) 0 (0)Diagnostic methods

surgery 1 2 16 1 2 9 cytology 1 3 2 0 3 4

clinical 15 16 0 3 15 6 Abbreviations: a, nonmucinous cysts; b, pseudocysts; c, nonmucious neoplastic cysts; d, mucinous cysts; EUS, endoscopic ultrasound. 2

on March 6, 2021. ©

2015 Am

erican Association for C

ancer Research.

cancerres.aacrjournals.org D

ownloaded from

Author m

anuscripts have been peer reviewed and accepted for publication but have not yet been edited.

Author M

anuscript Published O

nlineFirst on June 29, 2015; D

OI: 10.1158/0008-5472.C

AN

-14-2896


27

Table 2. MRM transitions of the representative peptides corresponding to 9 proteins

Description Peptide sequence Mass Information

Target ion Light (Q1/Q3) Heavy (Q1/Q3) CE(eV)

AFM AIPVTQYL*K 2/y5 516.8 / 652.4 520.3 / 659.4 28 2/y6 516.8 / 751.4 520.3 / 758.5 28 2/y7 516.8 / 848.5 520.3 / 855.5 28

CTRC VSAYIDWI*NEK 2/y5 669.3 / 689.4 672.8 / 696.4 34 2/y7 669.3 / 917.5 672.8 / 924.5 34 2/y8 669.3 / 1080.5 672.8 / 1087.6 34

CELA3A LANPVSL*TDK 2/y6 529.3 / 662.4 532.8 / 669.4 28 2/y7 529.3 / 759.4 532.8 / 766.4 28 2/y8 529.3 / 873.5 532.8 / 880.5 28

LGALS3BP IDITLSSV*K 2/y6 488.3 / 634.4 491.3 / 640.4 26 2/y7 488.3 / 747.5 491.3 / 753.5 26 2/y8 488.3 / 862.5 491.3 / 868.5 26

FCGBP GNPAVSYV*R 2/y5 481.8 / 623.3 484.8 / 629.4 26 2/y6 481.8 / 694.4 484.8 / 700.4 26 2/y7 481.8 / 791.4 484.8 / 797.5 26

REG1A WHWSSGSLV*SYK 2/y10 718.8 / 1113.6 721.9 / 1119.6 37 2/y8 718.8 / 840.5 721.9 / 846.5 37 2/y9 718.8 / 927.5 721.9 / 933.5 37

LCN2 WYVVGLAGNAIL*R 2/y10 716.4 / 983.6 719.9 / 990.6 37 2/y11 716.4 / 1082.7 719.9 / 1089.7 37 2/y9 716.4 / 884.5 719.9 / 891.5 37

AMY2B LTGLLDLAL*EK 2/y7 593.4 / 801.5 596.9 / 808.5 27 2/y8 593.4 / 914.6 596.9 / 921.6 27 2/y9 593.4 / 971.6 596.9 / 978.6 27

PIGR IIEGEPNL*K 2/y6 506.8 / 657.4 510.3 / 664.4 27 2/y7 506.8 / 786.4 510.3 / 793.4 27 2/y8 506.8 / 899.5 510.3 / 906.5 27

NOTE: Endogenous peptides (light) and isotope-labeled peptides (heavy) were chosen to measure biomarkers of interest. * represents an amino acid labeled with a heavy isotope, 13C15N. The transitions (Q1/Q3) were optimized using MRMPilot software. Abbreviations: AFM, afamin; CTRC, chymotrypsin C (caldecrin); CELA3A, chymotrypsin-like elastase family, member 3A ; LGALS3BP, lectin, galactoside-binding, soluble, 3 binding protein; FCGBP, Fc fragment of IgG binding protein; REG1A, lithostathine-1-alpha; LCN2, lipocalin 2; AMY2B, amylase, alpha 2B (pancreatic); PIGR, polymeric immunoglobulin receptor.




28

Table 3. A comparison of potential proteomic markers 1

Marker (amol/uL) Median (95% CI) P-value AUC

Validation cohort 1 NMC (n = 38)

MC (n = 18) NMC vs. MC

NMC2 vs. MC

NMC vs. MC

NMC2 vs. MC NMC1 (n = 17) NMC2 (n = 21)

AFM 261 (139 - 724) 241 (171 - 577) 41.0 (20 - 77) 0.000 0.000 0.825 0.837

CTRC 374 (207 - 1717) 92 (44 - 115) 354 (95 - 904) 0.188 0.011 0.608 0.73

CELA3A 253 (76 - 996) 31 (22 – 186) 287 (139 - 827) 0.162 0.035 0.617 0.69

LGALS3BP 94 (45 - 157) 44 (22 - 87) 41 (14 - 58) 0.149 0.681 0.635 0.559

FCGBP 1149 (617 - 3273) 223 (119 - 418) 1273 (701 - 4431) 0.021 0.001 0.696 0.715

REG1A 321 (51 - 688) 59 (39 - 153) 490 (148 - 1107) 0.017 0.001 0.727 0.815

LCN2 1414 (470 - 4755) 218 (95 - 608) 1717 (649 - 5910) 0.067 0.004 0.719 0.771

AMY2B 1680 (251 - 4928) 88 (39 - 380) 1304 (668 - 5446) 0.074 0.003 0.652 0.731

PIGR 13 (9 - 45) 32.5 (11 - 93) 89 (29 - 609) 0.004 0.065 0.728 0.664

Validation cohort 2 NMC (n = 24)

MC (n = 19) NMC vs. MC

NMC2 vs. MC

NMC vs. MC

NMC2 vs. MC NMC1 (n = 4) NMC2 (n = 20)

AFM 304 (187 - 510) 48 (21 - 77) 0.000 0.000 0.839 0.827

REG1A 178 (109 - 194) 9709 (7631 - 28600) 0.000 0.000 0.945 0.906

on March 6, 2021. ©

2015 Am


ancer Research.


ownloaded from

Author m


Author M



OI: 10.1158/0008-5472.C

AN

-14-2896


29

LCN2 51 (41-89) 1130 (441 - 422) 0.000 0.000 0.895 0.882

PIGR 29 (22-55) 920 (346 - 3032) 0.000 0.000 0.887 0.897

NOTE: Statistical comparisons were made among the three groups of pancreatic cysts for the top 9 markers (NMC [controls] vs. MC [cases] or NMC2 1

[controls] vs. MC [cases]). A P-value ≤ 0.05 was considered statistically significant. 2

Abbreviations: CI, confidence interval; AUC, the area under the curve. 3

4

5

6

on March 6, 2021. ©

2015 Am


ancer Research.


ownloaded from

Author m


Author M



OI: 10.1158/0008-5472.C

AN

-14-2896














Published OnlineFirst June 29, 2015.Cancer Res Jisook Park, Hwan Sic Yun, Kwang Hyuck Lee, et al. mucinous and nonmucinous pancreatic cystsDiscovery and Validation of Biomarkers that distinguish

Updated version

10.1158/0008-5472.CAN-14-2896doi:

Access the most recent version of this article at:

Material

Supplementary

http://cancerres.aacrjournals.org/content/suppl/2021/03/03/0008-5472.CAN-14-2896.DC1

Access the most recent supplemental material at:

Manuscript

Authoredited. Author manuscripts have been peer reviewed and accepted for publication but have not yet been

E-mail alerts related to this article or journal.Sign up to receive free email-alerts

Subscriptions

Reprints and

[email protected] at

To order reprints of this article or to subscribe to the journal, contact the AACR Publications

Permissions

Rightslink site. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC)

.http://cancerres.aacrjournals.org/content/early/2015/06/27/0008-5472.CAN-14-2896To request permission to re-use all or part of this article, use this link



http://cancerres.aacrjournals.org/lookup/doi/10.1158/0008-5472.CAN-14-2896

http://cancerres.aacrjournals.org/content/suppl/2021/03/03/0008-5472.CAN-14-2896.DC1

http://cancerres.aacrjournals.org/cgi/alerts

mailto:[email protected]

http://cancerres.aacrjournals.org/content/early/2015/06/27/0008-5472.CAN-14-2896


Documents

RESEARCH ARTICLE Discovery and Validation of Biomarkers that … · 2015/6/27 · 81 of pancreatic cysts in clinical practice, the diagnostic accuracy of these techniques is not