Virchows Arch DOI 10.1007/s00428-017-2114-2
ORIGINAL ARTICLE
Annexin A10 optimally differentiates between intrahepatic cholangiocarcinoma and hepatic metastases of pancreatic ductal adenocarcinoma: a comparative study of immunohistochemical markers and panels Julia Kälsch 1,2 & Juliet Padden 3 & Stefanie Bertram 1 & Leona L. Pott 1,3 & Henning Reis 1 & Daniela Westerwick 1 & Christoph M. Schaefer 1 & Jan-P. Sowa 2 & Dorothe Möllmann 1 & Christian Fingas 4 & Alexander Dech ne 2 & Barbara Sitek 3 & Martin Eisenacher 3 & Ali Canbay 2 & Maike Ahrens 3 & Hideo A. Baba 1
Received: 23 August 2016 / Revised: 12 January 2017 / Accepted: 21 March 2017 # Springer-Verlag Berlin Heidelberg 2017
Abstract Discriminating intrahepatic cholangiocarcinoma (ICC) from hepatic metastases of pancreatic ductal adenocarcinoma (mPDAC) can be challenging. While pathologists might depend on clinical information regarding a primary tumor, their diagnosis will lead the patient either to potentially curative surgery (for ICC) or to palliation (for mPDAC). Beyond the validation of recently published potential biomarkers for PDAC (primary or metastatic) in a large cohort, we assessed diagnostic performance of the most promising candidates in the challenging task of discriminating metastatic PDAC (mPDAC) from ICC. In a training set of 87 ICC and 88 pPDAC, our previously identified biomarkers Annexin A1 (ANXA1), ANXA10, and ANXA13 were tested and compared with 11 published biomarkers or panels (MUCIN 1,
Agrin, S100P, MUC5 AC, Laminin, VHL, CK 17, N-Cadherin, ELAC2, PODXL and HSPG2). Biomarkers with best results were further tested in an independent series of biopsies of 27 ICC and 36 mPDAC. Highest AUC values (between 0.72 and 0.84) for the discrimination between ICC and pPDAC were found in the training set for Annexin A1, Annexin A10, MUC5 AC, CK17, and N-Cadherin. These markers were further tested on an independent series of liver biopsies containing ICC or mPDAC. Diagnostic characteristics were evaluated for individual markers as well as for 3× panels. ANXA 10 showed the highest diagnostic potential of all single markers, correctly classifying 75% of mPDAC and 85% of ICC. Our results suggest that ANXA10 may be useful to differentiate between ICC and mPDAC, when only a tissue specimen is available.
Maike Ahrens and Hideo A. Baba contributed equally to the study. Electronic supplementary material The online version of this article (doi:10.1007/s00428-017-2114-2) contains supplementary material, which is available to authorized users. * Hideo A. Baba
[email protected] 1
Institute of Pathology, University Hospital of Essen, University of Duisburg-Essen, Hufelandstr. 55, 45147 Essen, Germany
2
Department of Gastroenterology and Hepatology, University Hospital of Essen, University of Duisburg-Essen, Hufelandstr. 55, 45147 Essen, Germany
3
Medizinisches Proteom-Center, Ruhr-Universität Bochum, Universitätsstr 150, 44780 Bochum, Germany
4
Department of General, Visceral and Transplantation Surgery, University Hospital of Essen, University of Duisburg-Essen, Hufelandstr. 55, 45147 Essen, Germany
Keywords Intrahepatic cholangiocarcinoma . Ductal pancreatic adenocarcinoma . Biomarker . Annexin A1 . Annexin A10
Introduction The term malignant liver tumor covers a broad spectrum of underlying tumor types of different origin. Intrahepatic cholangiocarcinoma (ICC), primary adenocarcinoma of the liver originating from bile ducts, is second in frequency after hepatocellular carcinoma, which is the most frequent primary liver cancer [1–3]. Whereas a diagnosis of hepatocellular carcinoma can mostly be made on histomorphological criteria and immunohistochemical markers, the differentiation between ICC and
Virchows Arch
metastasis of pancreatic ductal adenocarcinoma (mPDAC) can be challenging. Histologically, ICC and mPDAC are very similar, both containing ductular structures composed of epithelial cells with nuclear atypia and desmoplastic stroma. In particular, when only a small needle core biopsy is available, clinical information is indispensable to make a final diagnosis. The distinction is important, as treatment of these two cancers is significantly different; patients with early-stage ICC may undergo curative surgical resection while treatment options for patients with mPDAC are limited to palliative chemotherapy. Clinical information might not be available, in particular when a biopsy of the hepatic mass is taken prior to appropriate imaging studies or submitted for consultation from an external institution without appropriate clinical information. To distinguish ICC from mPDAC, various IHC markers have been studied with limited or no success. We previously identified in a proteomic approach annexins ANXA1, ANXA10, and ANXA13 to be most promising for the distinction between ICC and primary PDAC. For this study, we combined these with the most promising published markers (MUCIN 1, Agrin, S100P, MUC5 AC, Laminin, VHL, CK 17, N-Cadherin, ELAC2, PODXL, and HSPG2) for use on a training set of 87 ICC and 88 primary PDAC cases [4–11]. The five best-performing markers were then applied to an independent validation set consisting of core biopsies of 27 ICC and 36 PDAC liver metastases, in an approach mimicking the challenging situation in clinical practice.
Materials and methods Study design The study incorporates two independent case series as summarized in supplementary Table 1. We included tissue samples from a previous study. Training set The training set consisted of tissue samples of 87 cases of ICC and of 88 cases of primary pancreatic cancer. This was used to assess the quality of the available antibodies and performance of the 14 candidate immunohistochemical markers based on ROC curve analysis of immunoreactivity scores (IRS). Immunohistochemistry was performed using tissue microarrays. The five most promising markers were then applied to a second case series of ICC and mPDA biopsies.
Diagnoses in these cases were established in a multidisciplinary approach, including imaging results. In addition to the IRS, which is commonly used for study purposes, we also included an approach based upon global visual perception, which corresponds more to daily clinical practice.
Tissue microarray construction and immunohistochemistry ICC tumor tissue as well as corresponding tumor-free liver tissue for the training set was collected during surgery at University Hospital of Essen and University Hospital of Tübingen, Germany. All ICC cases were intrahepatic, as they most commonly are [12, 13]. Diagnoses were made in accordance with the recent WHO classification [14]. Of PDAC, primary tumor tissue was collected during surgery at the University Hospital of Essen, whereas metastatic tissue was collected via guided biopsy at the Department of Gastroenterology and Hepatology, University Hospital of Essen. All ICC and PDAC were moderate to poorly differentiated adenocarcinomas (grade G2 or G3). The study protocol conforms to guidelines of the 1975 Declaration of Helsinki. The IHC markers were evaluated by immunohistochemistry using tissue microarrays (TMAs) for the training set. Technical details are given in supplementary Table 2. Immunohistochemical staining and the subsequent light microscopic study were performed as described previously [10]. To evaluate the immunohistochemical stains, two independent observers used a semi-quantitative scoring system, analogous to the IRS established by Remmele and Stegner, as described in our previous study [15]. The score represents the product of quantity of stained tumor tissue (0% 0 points, 1–5%1 point, 6–10% 2 points, 11–50% 3 points, >50% 4 points) and staining intensity (none 0 points, weak 1 point, moderate 2 points, deeply stained 3 points). The final score ranges from 0 to 12 points. When scores were discordant, a third observer determined a final IRS. Discordant cases had slightly different values for staining intensity and/or percentage of stained cells between the two observers. Differences of more than one level in either quantity of stained tumor tissue and/or staining intensity did not occur. For the validation set of core biopsies, all three observers submitted an IRS; the most experienced pathologist (HB) provided the final diagnosis.
Validation set
Statistical analysis
To emulate the diagnostic challenge in clinical practice, the five selected markers were applied to an independent set of biopsies of 36 liver metastases of PDAC and 27 primary ICC.
Statistical analysis was conducted in R (R Foundation for Statistical Computing, Vienna, Austria, version 3.2.3) and SPSS (V22; IBM, Armonk, NY, USA).
Virchows Arch
Correspondence between the immunohistochemical staining results of the two case sets was assessed applying the two-sided Wilcoxon test to the IRS for each marker. The resulting p values were adjusted for multiple testing according to Benjamini and Hochberg, to reduce the false discovery rate. Differences were considered significant for (adjusted) p values <0.05. The overall discriminative power of each candidate marker was assessed by means of the area under the receiver operating characteristics (ROC) curve (AUC). To assess diagnostic accuracy for a diagnosis of PDAC, we defined PDAC samples (either primary or metastases) as cases and ICC samples as controls. Thus, the sensitivity of a marker refers to the true probability of classifying a PDAC sample correctly, whereas specificity is the probability of excluding a PDAC, which corresponds to classifying an ICC sample correctly. Note that depending on the marker, either high or low IRS values may result in the assignment of ICC for example. For each marker and comparison, the optimal IRS cut-off value was chosen to maximize the sum of sensitivity and specificity (Youden criterion). For all ROC analyses, the R package pROC was used. For the independent clinical condition set, the following additional statistical methods were applied. Boxplots were provided for individual markers, as well as a heatmap for visualization of IRS data (Figs. 1 and 2). The samples (i.e., rows in the heatmap) are sorted by entity, while the columns are sorted according to hierarchical clustering. In addition to assessing the discriminative power of single markers, we also considered three marker panels, in order to further increase diagnostic accuracy. Using cut-off values derived from the univariate models, a three-marker panel classifies a sample as mPDAC if at least two of the individual three markers do so. We also developed an additional classification model to address possible inconclusive cases, in which only a small percentage of cells showed any staining. For inconclusive cases, application of a different marker might be helpful. This hierarchical approach appears less pertinent for a practicing pathologist and is detailed in the supplementary material (supplementary Fig. 1 and supplementary Table 3).
Results
biomarkers. Detailed information on the staining pattern and statistical outcome of these biomarkers is described in supplementary Table 4. From training to clinical conditions—Evaluation of the best biomarkers In general, expression of the different biomarkers in the clinical validation set was similar to that in the training set (see Fig. 2). However, to account for possible differences between the expression patterns in primary and metastatic PDAC, cut-offs were optimized in the clinical validation set in analogy to the training set (see Table 1). In addition to antibody’s performance and IRS cut-off in the clinical validation set, the following staining patterns were observed. Annexin A1 In PDAC, ANXA1 showed moderate to marked cytoplasmic staining, whereas nuclei stained weakly. ICC showed negative to moderate cytoplasmic staining intensity of malignant cells, while the stroma was very frequently stained weakly to moderately. At IRS >3, 92% of mPDACs were classified correctly (sensitivity) whereas this was 56% for ICC (specificity) (p < 0.05). Annexin A10 Similar to ANXA1, ANXA10 stained significantly more pPDAC than ICC cells with a diffuse cytoplasmatic pattern. The staining intensity ranging from moderate to strong was also observed for nuclei. At IRS >1, 75% of mPDACs were classified correctly (sensitivity) whereas this was 85% for ICC (specificity) (p < 0.05). MUC5 AC Cytoplasmic moderate to strong expression was seen in pPDAC significantly more often than in ICC which was weakly if at all stained in most cases. Both stroma and nuclei remained unstained. At IRS >7, 77% of mPDACs were classified correctly (sensitivity) whereas this was 78% for ICC (specificity) (p < 0.05).
Expression of antibodies in the training set CK17 Comparing ICC and pPDAC in the training set, the five best biomarkers turned out with both estimated AUC (values reported in the following) and the lower limit of the AUC CI above 0.7: ANXA1 (0.89), ANXA10 (0.81), MUC5 AC (0.82), CK17 (0.79), N-Cadherin (0.79). All yielded an adjusted p value below 0.05 in the Wilcoxon test. Supplementary Fig. 2 shows representative photomicrographs of these five
Again, PDACs were stained more frequently, only cytoplasmatic, without nuclear staining. At IRS >2, in the training set, 80% of mPDACs were classified correctly (sensitivity) whereas this was 72% for ICC (specificity) (p < 0.05). In the validation set, this was 53% (specificity) and 82% (sensitivity).
Virchows Arch Fig. 1 Heatmap: visualization of the best five markers’ performance in clinical condition set
N-cadherin Immunohistochemical analysis of N-cadherin revealed significantly more staining in ICC compared to PDAC, mostly in moderate intensity. Mostly, membranous staining was found in addition to cytoplasmic staining. At IRS <11, 100% of mPDACs were classified correctly (sensitivity) whereas this was 27% for ICC (specificity) (p = 0.12). In summary, ANXA A10 showed the best performance for the discrimination between ICC and mPDAC when used as a single marker. Supplementary Table 5 shows the distribution of IRS in mPDAC. In an attempt to construct a diagnostic tool with even higher accuracy, we also tested three marker panels. All ten possible combinations classified >70% of mPDAC correctly. The best results were achieved by a combination of ANXA1 with ANXA10 and CK17. Representative photomicrographs of this three-marker panel are shown in Fig. 3. When two out of three biomarkers were positive, this panel
diagnosed 83% of mPDAC correctly (sensitivity) while this was 85% for ICC, resulting in an accuracy of 84%. Overall, adding ANXA10 to a panel together with ANXA1 and CK17 did not improve specificity compared to ANXA10 alone (85%). The increase in sensitivity is considered small and it is questionable whether this justifies the use of a three-marker panel.
Discussion Pathologists still need clinical information, notably diagnostic imaging results, to make a definitive diagnosis of ICC or metastatic PDAC because of their morphological resemblance. Immunohistochemical biomarkers may support the differentiation between ICC and mPDAC. Various staining protocols and antibodies were used in published studies but they did not improve either sensitivity or specificity [5, 6, 8, 11]. In an
Virchows Arch
Fig. 2 Boxplots of performance in ICC and PDAC for a ANXA 1, b ANXA 10, c MUC5 AC, d CK17, and e N-cadherin in training and clinical condition set
attempt to increase diagnostic accuracy, we tested a combination of earlier published markers and Annexins, which we had identified in proteomic studies. The best performance was obtained by ANXA10 as marker for an IRS >1 with a specificity of 85% and sensitivity of 75% in univariate analysis. Although statistical evaluation supports an IRS cut-off >1 for ANXA10, application of this threshold in daily practice might be hazardous as this corresponds to only a few stained cells or a very low staining intensity, which may not be very convincing. For clinical practice, a minimum of 10% stained tumor cells might provide a more solid basis for definitive
diagnosis, as was proposed earlier by Hooper et al. [6]. Of note, our IRS analyses were based on all samples in the study, in order to make our results comparable with those of earlier studies. As the amount of tissue in the case of a biopsy of a metastatic lesion is usually limited, only a small number of mPDAC cases could be evaluated. Macroscopically, ICCs present in three different forms: the mass-forming type, the periductal infiltrating type, and the intraductal-growth type. All of our ICCs were of the mass-forming type or a combination of mass forming and periductal infiltrating types. None of
Table 1 Characteristics of univariate biomarkers comparing intrahepatic cholangiocarcinoma (ICC) tumor samples to metastatic pancreatic ductal adenocarcinoma (mPDAC) within the clinical condition set
Annexin A1 Annexin A10 MUC5 AC CK 17 N-cadherin
AUC
CI lower, upper
p value
Classified as mPDAC if
Percentage of correctly classified ICC × Specificity
Percentage of correctly classified PDAC × Sensitivity
0.75 0.830 0.795 0.634 0.59
0.619, 0.881 0.733, 0.926 0.682, 0.907 0.499, 0.769 0.433, 0.746
0.001 0.000 0.000 0.077 0.230
>3 >1 >7 >2 <10
55.6 85.2 77.8 81.5 26.9
91.7 75.0 76.5 52.8 100
Virchows Arch Fig. 3 Triple-panel (ANXA1, ANXA10, and CK17) in ICC and mPDAC (all images original magnification ×400)
these tumors showed mucinous differentiation, which is usually associated with the intraductal-growth type. We therefore did perform sub-classification of ICC based on expressed mucin types. Marker studies in PDAC and ICC have been performed before. Notably, Gandou et al. studied the transcription factors HES and PDX1 in PDAC, ICC, hilar cholangiocarcinoma, and combined hepatocellular cholangiocarcinoma, in an attempt to categorize cholangiocarcinomas into those related to hepatic progenitor cells or bile ductules and those related to large bile ducts [16]. They found both markers strongly expressed in PDAC, to a lesser degree in hilar cholangiocarcinoma but only weakly in ICC. The authors related these expression patterns to differences in expression of these transcription factors during embryonic development. Other authors studied biomarker panels to distinguish between ICC and PDAC but mostly comparing primary PDAC to ICC. Only three studies focused on mPDAC [6, 8, 11]. Lok et al. applied a biomarker panel of S100, pVHL, MUC5AC, and CK17 to a cohort of 41 ICC and 60 primary PDAC. They did not include metastatic PDAC, and none of these markers when used alone allowed determination of tumor origin [7]. Chu et al. combined CK17 and MUC1 and found a positive predictive value for a diagnosis of pPDAC in 76% of 46 cases [5]. Goldstein et al. found Cytokeratin 17 staining in 66% of
35 pPDAC, taking all staining intensities together [17]. McKinney et al. found a significant difference in protein expression between pPDAC and mPDAC cell lines [18]. Hooper et al. [6]. tested a biomarker panel consisting of ELAC2 (HPC2) and N-cadherin to distinguish between ICC and metastatic PDAC. They achieved a sensitivity of 70% with a specificity of 68%, regardless of whether the mPDAC were hepatic or peritoneal. N-Cadherin alone correctly classified 58% of cases of primary ICC and 22% of cases of mPDAC, which according to our definition results in a sensitivity of 58% and a specificity of 78%. Applied to our cohort, N-cadherin showed a sensitivity of 100% but specificity of only 27%. The difference between these results cannot be related to antibody specificity, as we used the same anti-N-cadherin antibody. In our study, ANXA10 performed better than did ELAC2 and N-Cadherin. Somoracz et al. tested the expression of Agrin in mPDAC and ICC but focused on immunoreactivity of microvessels versus basement membranes rather than on expression patterns in tumor cells [11]. Only one other study has tested ANXA10 performance in metastatic PDAC. Lu et al. studied ANXA10 as an immunohistochemical marker to determine origin in case of adenocarcinomas of the gastrointestinal tract and pancreatobiliary system of unknown primary site [8]. ANXA10 stain was positive in metastatic gastric and colorectal cancer and in 83% of
Virchows Arch
mPDAC. This result is close to ours (85%), with a slight difference in sensitivity as only 17% intrahepatic cholangiocarcinomas were stained (25% in our study). The authors did not specifically address the differential diagnosis between ICC and mPDAC nor did they use an IRS or propose a cut-off to call a sample positive. ANXA10 protein expression has been found in human carcinomas of various anatomic sites (stomach, urinary bladder, urothel, or oral cavity) [8, 19]. Most patients with a liver metastasis will have undergone endoscopy of the gastrointestinal tract to exclude, e.g., gastric cancer, before liver biopsy is performed. If such information is not available, an ANXA10 positive lesion cannot be diagnosed as metastatic PDAC. The performance of ANXA10 as a marker to diagnose organ of origin other than pancreas needs further study. We conclude that ANXA10 is the best immunohistochemical protein marker to discriminate between ICC and mPDAC. For daily practice, we propose a threshold of 10% ANXA10-stained tumor cells to make a diagnosis of mPDAC, based on literature data [6]. The IRS threshold of >1 has high diagnostic discriminative power but cannot be reliably applied in daily diagnostic practice. Acknowledgement The authors thank Laura Malkus who provided expertise that greatly assisted the research.
5.
6.
7.
8.
9.
10.
11.
Compliance with ethical standards Ethics Statement The study protocol conforms to the ethical guidelines of the 1975 Declaration of Helsinki. Conflict of interest The authors declare that they have no conflict of interest. Funding This study was supported by PROFILE consortium Ruhr, European Regional Development Fund (ERDF) objective 2 Bregional Competitiveness and Employment^ 2007–2013 North-Rhine-Westphalia (Germany) and by the de.NBI project (FKZ 031 A 534A) of the Bundesministerium für Bildung und Forschung (BMBF). JK received intramural research funds of the Medical Faculty of the University of Duisburg-Essen (Interne Forschungsförderung Essen, IFORES).
References Bosch FX, Ribes J, Cleries R, Diaz M (2005) Epidemiology of hepatocellular carcinoma. Clinics in Liver Disease 9:191–211. doi:10.1016/j.cld.2004.12.009 2. Bosch FX, Ribes J, Diaz M, Cleries R (2004) Primary liver cancer: worldwide incidence and trends. Gastroenterology 127:S5–S16 3. Khan SA, Toledano MB, Taylor-Robinson SD (2008) Epidemiology, risk factors, and pathogenesis of cholangiocarcinoma HPB. the official journal of the International Hepato Pancreato Biliary Association 10:77–82. doi:10.1080/13651820801992641 4. Bai XF, Ni XG, Zhao P, Liu SM, Wang HX, Guo B, Zhou LP, Liu F, Zhang JS, Wang K, Xie YQ, Shao YF, Zhao XH (2004) Overexpression of annexin 1 in pancreatic cancer and its clinical significance. World J Gastroenterol 10:1466–1470
12. 13. 14.
15.
16.
17.
1.
18.
19.
Chu PG, Schwarz RE, Lau SK, Yen Y, Weiss LM (2005) Immunohistochemical staining in the diagnosis of pancreatobiliary and ampulla of Vater adenocarcinoma: application of CDX2, CK17, MUC1, and MUC2. Am J Surg Pathol 29:359–367 Hooper JE, Morgan TK, Grompe M, Sheppard BC, Troxell ML, Corless CL, Streeter PR (2012) The novel monoclonal antibody HPC2 and N-cadherin distinguish pancreatic ductal adenocarcinoma from cholangiocarcinoma. Hum Pathol 43:1583–1589. doi:10. 1016/j.humpath.2011.11.012 Lok T, Chen L, Lin F, Wang HL (2014) Immunohistochemical distinction between intrahepatic cholangiocarcinoma and pancreatic ductal adenocarcinoma. Hum Pathol 45:394–400. doi:10.1016/j. humpath.2013.10.004 Lu SH, Yuan RH, Chen YL, Hsu HC, Jeng YM (2013) Annexin A10 is an immunohistochemical marker for adenocarcinoma of the upper gastrointestinal tract and pancreatobiliary system. Histopathology 63:640–648. doi:10.1111/his.12229 Ney JT, Zhou H, Sipos B, Buttner R, Chen X, Kloppel G, Gutgemann I (2007) Podocalyxin-like protein 1 expression is useful to differentiate pancreatic ductal adenocarcinomas from adenocarcinomas of the biliary and gastrointestinal tracts. Hum Pathol 38: 359–364. doi:10.1016/j.humpath.2006.08.025 Padden J, Ahrens M, Kalsch J, Bertram S, Megger DA, Bracht T, Eisenacher M, Kocabayoglu P, Meyer HE, Sipos B, Baba HA, Sitek B (2015) Immunohistochemical markers distinguishing cholangiocellular carcinoma from pancreatic ductal adenocarcinoma discovered by proteomic analysis of microdissected cells. Molecular & cellular proteomics : MCP. doi: 10.1074/mcp.M115. 054585 Somoracz A, Tatrai P, Horvath G, Kiss A, Kupcsulik P, Kovalszky I, Schaff Z (2010) Agrin immunohistochemistry facilitates the determination of primary versus metastatic origin of liver carcinomas. Hum Pathol 41:1310–1319. doi:10.1016/j.humpath.2009.10.029 Lazaridis KN, Gores GJ (2005) Cholangiocarcinoma. Gastroenterology 128:1655–1667 Shaib Y, El-Serag HB (2004) The epidemiology of cholangiocarcinoma. Semin Liver Dis 24:115–125. doi:10.1055/s-2004-828889 Bosman FT, World Health Organization., International Agency for Research on Cancer (2010) WHO classification of tumours of the digestive system. Chapter 10. International Agency for Research on Cancer, Lyon Remmele W, Stegner HE (1987) Recommendation for uniform definition of an immunoreactive score (IRS) for immunohistochemical estrogen receptor detection (ER-ICA) in breast cancer tissue. Pathologe 8:138–140 Gandou C, Harada K, Sato Y, Igarashi S, Sasaki M, Ikeda H, Nakanuma Y (2013) Hilar cholangiocarcinoma and pancreatic ductal adenocarcinoma share similar histopathologies, immunophenotypes, and development-related molecules. Hum Pathol 44:811–821. doi:10.1016/j.humpath.2012.08.004 Goldstein NS, Bassi D (2001) Cytokeratins 7, 17, and 20 reactivity in pancreatic and ampulla of vater adenocarcinomas. Percentage of positivity and distribution is affected by the cut-point threshold. Am J Clin Pathol 115:695–702. doi:10.1309/1NCM-46QX-3B5T7XHR McKinney KQ, Lee JG, Sindram D, Russo MW, Han DK, Bonkovsky HL, Hwang SI (2012) Identification of differentially expressed proteins from primary versus metastatic pancreatic cancer cells using subcellular proteomics. Cancer Genomics Proteomics 9:257–263 Shimizu T, Kasamatsu A, Yamamoto A, Koike K, Ishige S, Takatori H, Sakamoto Y, Ogawara K, Shiiba M, Tanzawa H, Uzawa K (2012) Annexin A10 in human oral cancer: biomarker for tumoral growth via G1/S transition by targeting MAPK signaling pathways. PLoS One 7:e45510. doi:10.1371/journal.pone.0045510