Proteomics Profiling of Microdissected Low- and High-Grade Prostate Tumors Identifies Lamin A as a Discriminatory Biomarker Sergej Skvortsov,†,‡,§ Georg Scha¨fer,†,|,⊥ Taras Stasyk,‡ Christian Fuchsberger,⊥,¶ Guenther Karl Bonn,+ Georg Bartsch,⊥ Helmut Klocker,*,⊥,# and Lukas Alfons Huber*,‡,# Biocenter, Division of Cell Biology, Innsbruck Medical University, Innsbruck, Austria, Department of Therapeutic Radiology and Oncology, Innsbruck Medical University, Innsbruck, Austria, Institute of Pathology, Innsbruck Medical University, Innsbruck, Austria, Department of Urology, Division of Experimental Urology, Innsbruck Medical University, Innsbruck, Austria, Institute for Bioinformatics and Translational Research, UMIT, Hall in Tyrol, Austria, and Institute of Analytical Chemistry and Radiochemistry, Leopold-Franzens University, Innsbruck, Austria Received September 8, 2010
Proteomics screening methods for the identification of diagnostic and prognostic biomarkers in cancer are still lagging behind DNA- or RNA-based analysis. We used two-dimensional differential gel electrophoresis (2D-DIGE) in combination with laser capture microdissection (LCM) and MALDI-TOF/ TOF mass spectrometry to determine differentially abundant proteins and candidate biomarkers in prostate cancer. Paired (benign and tumor) samples were isolated from 23 Gleason Score 6 (GS 6) and 23 Gleason Score 8 and higher (GS 8+) radical prostatectomy specimens and subjected to 2D-DIGE analysis. Minimal fluorescent dye labeling was applied and electrophoresis performed with triple samples (paired benign and tumor; internal control) for each case of tumor. Nineteen differently abundant proteins were identified by mass spectrometry and further validated. One half of them were associated with glycolysis and the Warburg effect; these were upregulated in tumors. The upregulation correlated with tumor dedifferentiation and might be relevant for selection of therapeutic strategies. Among the other proteins, heat shock protein 60 (HSP60) was significantly upregulated in tumor tissue compared to its benign counterpart. Furthermore, lamin A was statistically highly discriminatory between low and high Gleason score tumors and might serve as a new biomarker of tumor differentiation and prognosis. Keywords: laser capture microdissection (LCM) • proteomics • biomarkers • HSP60 • lamin A
Introduction Prostate cancer is the most common malignancy in men and the second most common type of cancer death in men in Europe.1 Prostate specific antigen (PSA) level, Gleason score on biopsy, and clinical T-stage are currently the gold standard to assess the aggressiveness of prostate cancer at the time of its diagnosis and to enable decisions concerning the treatment.2 Although widespread use of PSA testing improved the detection * To whom correspondence should be addressed. Lukas A. Huber, Biocenter, Division of Cell Biology, Innsbruck Medical University, Fritz-Pregl Strasse 3, 6020 Innsbruck, Austria. Tel.: ++43 (0)512 9003 70170. Fax: ++43 (0)512 9003 73100. E-mail:
[email protected]. Helmut Klocker, Department of Urology, Division of Experimental Urology, Innsbruck Medical University, Anichstrasse 35, 6020 Innsbruck, Austria. Tel.: ++43 (0)512 504 24818. Fax: ++43 (0)512 504 24817. E-mail:
[email protected]. † These authors contributed equally to this work. ‡ Division of Cell Biology, Innsbruck Medical University. § Department of Therapeutic Radiology and Oncology, Innsbruck Medical University. | Institute of Pathology, Innsbruck Medical University. ⊥ Division of Experimental Urology, Innsbruck Medical University. ¶ UMIT. + Leopold-Franzens University. # These authors contributed equally to this work. 10.1021/pr100921j
2011 American Chemical Society
of organ-confined prostate tumors and allowed curative treatment by radical prostatectomy in an increasing number of patients, many men still develop advanced metastatic and deadly disease. For these cases, curative treatment options are currently not available. On the other hand, there is agreement that many patients are overtreated3 because reliable biomarkers to distinguish between aggressive tumors with metastatic potential and tumors that grow locally and have a low risk for metastatic progression are not available. In most cases, prostate tumors are multifocal, and the various parts of the tumor display different characteristics.4,5 The Gleason grading system that distinguishes between five histological patterns demonstrates a direct correlation with the outcome of the disease.6 However, at the time of diagnosis, the Gleason score of the whole tumor is not available and tumor characterization is based on histopathological examination of systematically taken needle biopsies, which may not be representative.7-9 Molecular features and biomarkers that define tumor phenotypes more precisely and improve the assessment of the aggressiveness of the tumor and the risk of progression are needed to provide a more specific treatment. This could avoid overtreatment and concomitant morbidity such as incontiJournal of Proteome Research 2011, 10, 259–268 259 Published on Web 10/26/2010
research articles
Figure 1. Study work flow.
nence and impotence. Other efforts concentrate on the identification of new targets and potential Achilles heals of highrisk tumors for developing better therapies for those cancer cases that have a high probability of progression to a deadly disease. Proteomics has become a major search engine for the discovery of diagnostic and prognostic cancer biomarkers,10,11 but the analysis of prostate cancer is challenging by its heterogeneity and the presence of various histological grades within the tumor.12 However, laser capture microdissection (LCM) provides an ideal method to extract tumor specimens and isolate cells with defined morphologies from the tissues.13 In the present study, as delineated in Figure 1, we characterized the protein profile of prostate carcinomas of low and high grades and of benign tissue using LCM in conjunction with 2D-DIGE, followed by MALDI-TOF-MS identification of differentially expressed proteins. Nineteen significantly regulated proteins were identified. Ten proteins belonged to protein classes that were clearly related to a classical Warburg effect. These proteins revealed highly significant differences between benign tissue and tumors, and also when low and high grade tumors were compared. Of the other nine proteins, lamin A revealed to be a promising marker to distinguish between lowand high-grade tumors, whereas HSP60 expression was the best marker to differentiate benign from malignant tissue.
Materials and Methods Tissue Specimens. Frozen and paraffin-embedded prostate tissue samples were obtained from previously untreated patients who had undergone radical prostatectomy after a tumor had been diagnosed during a PSA-based early detection program performed at the department of urology, Medical University of Innsbruck.14 The study had been approved by the ethics committee at the University of Innsbruck. Immediately after surgery, the prostate specimens were cooled and sent to the pathologist, who performed a rapid section and isolated a prostate slice that was embedded in Tissue-Tek OCT Compound (Sakura Finetek Germany GmbH, Staufen, Germany), snap-frozen in liquid nitrogen, and stored at -80 °C until use. The remainder of the prostate was fixed and paraffinembedded according to standard procedures. 260
Journal of Proteome Research • Vol. 10, No. 1, 2011
Skvortsov et al. Laser Capture Tissue Microdissection. Frozen sections of the tumor specimens were stained with hematoxylin and eosin for pathological analysis and exact localization of the tumor regions. Parallel unstained 8-µm frozen sections were used for laser capture microdissection (LCM). Slides were pretreated for 1 min each in 75% ethanol, 100% ethanol (twice), xylene (twice), and then air-dried. LCM was performed using a PixCell II LCM System (Arcturus Bioscience Inc., Mountain View, CA) with 5000 to 6000 laser pulses for each sample, corresponding to about 30 000 to 50 000 captured cells. Twenty-three tumor samples were isolated from a cohort of Gleason score 6 prostate tumors (Gleason pattern 3 + 3), and 23 samples from a second cohort of Gleason score 8 to 9 (Gleason pattern 4 + 4 or 4 + 5) tumors. Two samples each were taken from the 46 patients: one from the tumor and one from benign tissue, isolated from a region far apart from the tumor and without observable pathological changes. In addition from four cases stromal cells were isolated from the benign tissue and used as stromal control samples. After microdissection, the samples were solubilized in lysis solution (7 M urea, 2 M thiourea, 40 mM Tris base, 1% C7BZO, Sigma-Aldrich, St Louis, MO) and frozen at -80 °C until further processing. Two-Dimensional Differential in Gel Electrophoresis (2-D DIGE). Protein concentrations were determined with a commercial protein assay (BioRad Laboratories, Hercules, CA); 5 µg of proteins in 20 µL of lysis solution was labeled with 40 pmol of CyDye DIGE Fluor minimal dyes (GE Healthcare BioSciences AB, Uppsala, Sweden) [Cy3 or Cy5 for sample and Cy2 for the internal standard (a pool of equal amounts from all samples)] or with 3 nmol of saturation dyes (Cy3 for benign cells and Cy5 for tumor cells) for 30 min, resuspended in 280 µL of rehydration buffer (7 M urea, 2 M thiourea, 1% C7BZO, 0.5% IPG buffer, 60 mM DTT), and loaded on immobilized 18cm pH 3-11 NL gradient strips. For the first dimension analysis, active rehydration (50 V) was performed at 20 °C for 12 h. Isoelectric focusing was performed at 250 V for 30 min, 500 V for 1 h, 2000 V for 1 h and finally at 8000 V, until 32 000 V/hour had been achieved. For the second dimension analysis, samples were separated on 12.5% polyacrylamide gels with the Ettan Dalttwelve System according to the standard procedure recommended by the manufacturer (GE Healthcare Bio-Sciences AB, Uppsala, Sweden). After electrophoresis, gels were scanned using a Typhoon 94100 Imager at a resolution of 100 dpi (GE Healthcare BioSciences AB, Uppsala, Sweden). Identification of Proteins by Mass Spectrometry. Protein spots were excised from gels using an Ettan Spot Picker (GE Healthcare) and in-gel digested with modified trypsin, sequence grade (Promega, Madison, WI), as described elsewhere.15 The in-gel digests were concentrated and desalted with microZipTipC18 (Millipore, Billerica, MA). Peptides were eluted in acetonitrile solution containing R-cyano-4-hydroxycinnamic acid (Fluka, Buchs, Switzerland) as a matrix, and spotted directly onto the mass spectrometry target. Mass spectra were acquired using a MALDI-TOF/TOF Ultraflex instrument (Bruker Daltonics, Bremen, Germany). Flex Control 2.4 was used for data acquisition, and further data processing was performed with Flex analysis 2.4 and BioTools 2.2 software packages provided by the manufacturer of the mass spectrometer. Peptide mass fingerprintings (PMF) were interpreted against the human protein Swiss-Prot (Version 53.3, 274 295 sequences) and NCBInr (release date 20080924, 217 014 sequences) databases using MASCOT (http://www.matrixscience.com). The
research articles
Lamin A as a Discriminatory Biomarker search parameters used were: peptide mass tolerance of 0.1 Da, 1 missed cleavage, complete carbamidomethylation of cysteines and variable methionine oxidation. Immunohistochemistry. For validation of expression at the protein level, we used immunohistochemistry on paraffinembedded tissue from the same patient cohort. The paraffin slides directly corresponded to the frozen material used for LCM. Immunohistochemistry was performed with 3-µm paraffin tissue sections using the Ventana Discovery-XT staining automate and Ventana reagents (Roche, Basel Switzerland). For investigation of HSP60 immunoreactivity, standard CC2 pretreatment for antigen retrieval was followed by incubation with the primary antibody antiheat shock protein 60 (Clone LK1, Thermo Fisher Scientific, MA) diluted 1: 100 in antibody diluent for 1 h, and secondary universal antibody (Roche) solution for 30 min. The signal was visualized with the DAB map kit. The slides were then incubated with a second primary antibody solution of anti-p63 (Clone 4A4 + Y4A3, Thermo Fisher Scientific, 1:100 in antibody diluent) for 60 min and the secondary universal antibody for 30 min. This signal was visualized with the Ultra Map Kit using alkaline peroxidase and Fast Red as a substrate. Counter-staining was performed for 4 min with hematoxylin II and bluing reagent. For lamin A analysis, the standard CC2 pretreatment for antigen retrieval was followed by incubation with the primary antibody anti-lamin A (Clone 133A2, Abcam, 1: 150 in antibody diluent) for 1 h, and a secondary universal antibody solution for 30 min. The signal was visualized with the DAB map kit. Counter-staining was performed for 4 min with hematoxylin II and bluing reagent. The specificity of immunostaining was controlled by including an unspecific control antibody (DAKO Cytomation, Glostrup, Denmark). Immunoreactivity was scrutinized by an uropathologist (G.S.) and stratified according to the histology and Gleason patterns of the specimens using a 4 scale scoring system based on a combination of staining intensity and the number of positive cells: 0, no; 1, weak; 2, medium; and 3, high expression. For validation of expression in an independent set of tumor cases, a tissue microarray was used. It was composed of 94 × 4 cores of 94 tumor specimens obtained from patients undergoing radical prostatectomy. From each specimen 3 randomly taken cores of the tumor and 1 core of benign tissue were included in the tissue array. Statistical Analysis. Abundance of stained protein spots in the gels was analyzed by the use of the DeCyder DIA (Difference in-Gel Analysis) and DeCyder Biological Variation Analysis (BVA) software (GE Healthcare Bio-Sciences AB, Uppsala, Sweden). The statistical language R16 and packages from
Figure 2. Coupling of LCM and 2-D DIGE for assessing Gleason score-related protein expression (A) Laser capture microdissection images of frozen prostate cancer tissue sections before (a) and after LCM (b), and captured cancer cells (c). (B) Comparison of minimal and saturation labeling to optimize conditions for the best reproducibility, resolution and sensitivity for 2-D DIGE.
Bioconductor Project17 were used for all other analyses of the data, with one exception: the WEKA package18 was used to perform protein ranking. 2-D DIGE analysis: Two-sample t-statistics were used to identify differentially expressed proteins between respective groups. To correct for multiple testing, we chose to control the false discovery rate (FDR). According to Benjamini and Hochberg,19 FDR is defined by the anticipated number of erroneously rejected null hypotheses within the total number of rejected null hypotheses. Proteins with a FDR q-value < 0.5 were considered to be differentially expressed. Protein Ranking. To assess the importance of single proteins in discriminating between two groups a ranking was calculated using the information gain metric. The information gain of a
Table 1. Characteristics of Patientsa age [years]
pathological stage
Gleason score
PSA ( s.d. [ng/mL]
surgical margin
cohort
mean (range)
number
pT
number
GS
number
R
mean (Range)
Low risk cancer
61.8 (44-78)
6
17 6
0 1
5.0 ( 2.9 (1.7-13.0)
63.5 (46-83)
2a 2b 2c 2c 3a 3b 4
23
High risk cancer
5 1 17 5 10 5 3
13 10
8 9
14 9
0 1
8.9 ( 4.3 (2.5-17.3)
a Pathological staging (pT) and Gleason scoring (GS) was according to the TNM classification system, 6th edition 2002. Surgical margin (R) indicate no tumor at the surgical margin (0) or tumor detected at the surgical margin (1). PSA values are pre-surgery values.
Journal of Proteome Research • Vol. 10, No. 1, 2011 261
262
P04075
P48735
O75390
P00338
P36542
P06733
P18669
P40926
P22695
P29401
Isocitrate dehydrogenase
Citrate synthase
dehydrogenase A
L-Lactate
ATP synthase gamma chain
Alpha-enolase
Phosphoglycerate mutase 1
Malate dehydrogenase
Core protein II
Transketolase
accession no.
Fructose-bisphosphate aldolase A
target protein
Journal of Proteome Research • Vol. 10, No. 1, 2011 249
214
265
205
226
108
135
169
285
190
Mascot score
24 (40%)
19 (45%)
21 (61%)
14 (59%)
24 (53%)
13 (29%)
18 (42%)
15 (26%)
33 (53%)
14 (43%)
mass values matched/sequence coverage
Table 2. Function and Expression of Identified Proteins (Warburg Effect)
Glycolytic enzyme that catalyzes the reversible aldol cleavage or condensation of fructose-1,6-bisphosphate into dihydroxyacetone-phosphate and glyceraldehyde 3-phosphate Enzyme which participates in the tricarboxylic acid cycle Enzyme that catalyzes the first step in the tricarboxylic acid cycle Enzyme that catalyzes the interconversion of lactate and pyruvate Produces ATP from ADP in the presence of a proton gradient across the membrane. The gamma chain is believed to be important in regulating ATPase activity Multifunctional enzyme that plays a role in glycolysis and growth control Catalyzes the reversible reaction of 3-phosphoglycerate to 2-phosphoglycerate in the glycolytic pathway Enzyme in the tricarboxylic acid cycle that catalyzed the conversion of malate into oxaloacetate and involved in gluconeogenesis Component of the Cytochrome b-c1 complex, which is part of the mitochondrial respiratory chain Catalyzed reactions of the nonoxidative branch of the pentose-phosphate pathway, which provides the link between this pathway and glycolysis
function
1.31
1.19
1.64
1.56
1.48
1.28
1.40
1.18
1.09
1.15
regulation coefficient (Gleason 6 versus benign)
0.043
0.015
0.00027
0.018
0.041
0.021
0.035
0.18
0.26
0.27
student’s t test
1.59
1.61
1.64
1.65
1.67
1.7
1.75
1.76
1.79
1.85
regulation coefficient (Gleason 8+ versus Gleason 6)
0.00012
0.0024
0.0048
0.044
0.00096
0.0003
0.0027
0.0006
0.00095
0.0035
student’s t test
research articles Skvortsov et al.
P15309
P07288
Prostate-specific antigen
P20774
Mimecan
Prostatic acid phosphatase
P04264
Cytokeratin-1
P07355
O75955
Flotillin 1
Annexin A2
P02545
Lamin A
P62195
P10809
60 kDa heat shock protein
26S protease regulatory subunit 8
accession no.
target protein
172
313
174
176
193
207
224
312
190
Mascot score
Table 3. Function and Expression of Identified Proteins
19 (49%)
24 (32%)
18 (52%)
21 (48%)
16 (30%)
23 (44%)
28 (58%)
41 (56%)
25 (42%)
mass values matched/sequence coverage
Implicated in mitochondrial protein import and macromolecular assembly. Also occurs in the cytosol and the cell surface, participates in key intracellular pathways and in extracellular molecular interactions Lamins are components of the nuclear lamina which is thought to provide a framework for the nuclear envelope and may also interact with chromatin Act as a scaffolding protein within caveolar membranes, functionally participating in formation of caveolae or caveolae-like vesicles Integrin-binding protein modulating adhesion and cell survival signaling Belongs to a family of small leucine-rich proteoglycans, important for cellular growth, differentiation, and migration Involved in the ATP-dependent degradation of ubiquitinated proteins Multifunctional calcium-regulated protein implicated in a number of intra- and extracellular functions Enzyme produced by the prostate gland, released into the blood by dysfunction Produced in the ducts of the prostate gland and is a part of semen that causes it to liquefy.
function
0.91
-1.08
0.00056
0.84
-1.15
0.047
-2.85
1.29
0.082
0.73
-1.09
1.14
0.42
0.037
0.00062
student’s t test
1.07
-1.52
1.83
regulation coefficient (Gleason 6 versus benign)
-3.04
-1.84
1.46
1.47
1.58
1.65
1.85
2.08
2.13
regulation coefficient Gleason (8+ versus Gleason 6)
0.000025
0.012
0.0058
0.00084
0.0023
0.0011
0.00022
0.0003
0.00035
student’s t test
Lamin A as a Discriminatory Biomarker
research articles
Journal of Proteome Research • Vol. 10, No. 1, 2011 263
research articles protein with respect to discriminate two groups of samples, for example, benign and tumor samples, is the reduction in uncertainty when the protein expression is known. Proteins were ranked on the basis of their information gain value for accurate group separation; a lower rank means greater discriminating power. The robustness of the ranking was evaluated by 10-fold cross-validation in which the data were randomly divided into 10 subsamples. One subsample was used for validation; the remaining 9 subsamples were used for training. This process was repeated 10 times until each sample was used exactly once for validation. Calculation of final and average ranks, and standard deviations have been described previously.20 ROC Analysis. Receiver operating characteristic (ROC) curves21,22 were used to visualize the diagnostic performance of a protein. In brief, ROC curves are a graphic presentation of the relationship between sensitivity and 1-specificity for a binary classifier. The area under the curve (AUC) was calculated to determine the overall diagnostic accuracy. Heat Map Analysis. Heat maps were used to look for similarities between protein expression patterns within the three classes (benign, Gleason score 6, Gleason score 8+). Each colored cell on the two-dimensional map represents the protein expression value of the sample. Cells with a value of 0 are colored black. Increasingly positive values are indicated by reds of increasing intensity, and increasingly negative values by greens of increasing intensity, whereas white cells indicate missing values. Hierarchical clustering was used to arrange protein expression profiles and samples on the basis of their similarity. Dendrograms (trees) on the left and the top of the heat map show the clustering result; tree branches represent clusters obtained on each step of hierarchical clustering.
Results Experimental Design and Coupling LCM and 2-D DIGE for Analysis of Gleason-Associated Protein Expression Signatures. Prostate cancer often is a multifocal, heterogeneous tumor, being surrounded by benign tissue. Samples of high purity are needed for comparative profiling analyses. We used LCM for collection of malignant and benign cell populations from prostate tissue of 46 prostate cancer patients with lowergrade Gleason score 6 (GS6) and higher-grade Gleason score 8 and 9 (GS8+) tumors (Table 1). LCM was performed on frozen sections (Figure 2A) in areas carefully designated by the pathologist, using hematoxylin-eosin-stained parallel sections. Stromal cells were isolated as controls and examined for protein expression by the same procedure in order to exclude stromaassociated proteins during statistical evaluation of the protein profile in benign and malignant cells. After LCM and sample solubilization, we compared minimal and saturation labeling for 2-D DIGE to determine conditions for optimal reproducibility and resolution for coupling LCM and 2-D DIGE (Figure 2B). Minimal labeling of LCM tissue samples followed by 2-D DIGE revealed roughly the same number of spots as saturation labeling. This observation prompted us to combine LCM with minimal labeling. The advantage of this approach compared to saturation labeling is that an internal standard can be included into each sample, which enables the investigator to normalize protein expression in all studied samples for statistical analysis. Profile of Differentially Expressed Proteins. In order to identify differences in protein expression between benign, Gleason score 6 and 8+ tumor tissues, we determined protein 264
Journal of Proteome Research • Vol. 10, No. 1, 2011
Skvortsov et al. profiling in cell extracts obtained by laser capture microdissection of tissue sections from 46 patients by 2-D DIGE. Mean relative spot volumes, standard deviations, and differences in expression were calculated using DeCyder Biological Variation Analysis software; significantly differentially expressed proteins were identified on the basis of a FDR q-value < 0.05. When the expression level of individual proteins differed by a ratio of more than 1.5, protein spots were isolated, in-gel digested with trypsin, and analyzed using mass spectrometry followed by Swiss-Prot database search with MASCOT to assign identities and accession numbers (Tables 2 and 3). Functional Annotation of Differentially Expressed Proteins. We found 19 significantly and differentially expressed proteins. With reference to their assigned functions, these proteins might be divided into two groups: the first group included proteins related to the Warburg effect while the second group consisted of proteins allocated to various structural or functional families. All of these proteins are summarized in Tables 2 and 3. Figure 3 shows the expression profiles of the 19 differentially expressed proteins across the patients using hierarchical clustering. Characterization of Determined Proteins According to Their Discriminatory Power. The proteins identified in our comparative screen varied in their ability to assign samples to one of the groups. In order to decide which of them are best suitable as potential biomarkers to discriminate benign from tumor tissue and distinguish between different Gleason score tissues, we ranked the proteins based on a 10-fold cross validation procedure and subsequently evaluated them based on a ROC curve analyses. Ranking of the proteins according to their discriminatory power revealed that the 60 kDa heat shock protein (HSP60) occupied the first rank within the groups of benign/cancer, benign/GS6 and benign/GS8+ (Table 4). The AUC value for HSP60 was 0.89 when benign samples were compared to cancer samples (Figure 4A). Thus, HSP60 may serve as a good discriminator between cancer and noncancerous tissue. Interestingly, with regard to discrimination of low and high Gleason score tumors, HSP60 corresponded to the rank position 12 only (Table 4). However, lamin A was a mirror image of HSP60. While lamin A was found at rank position 1 with an AUC of 0.88 (Figure 4B) when comparing GS6 with GS8+ tissues, it occupied the last rank for distinguishing between tumor and benign tissue. Thus it is the best marker for discriminating low and high Gleason tumors but is of less value for discriminating tumor from nontumor samples. Immunohistochemical Analysis of HSP60 and Lamin A Protein Expression in Benign and Malignant Tissue. Next, the expression patterns of HSP60 and lamin A were confirmed using immunohistochemical analysis. Double-staining immunohistochemistry for HSP60 and the basal cell marker p63 showed weak or no expression of HSP60 in benign prostate glands (Figure 5A). P63 staining permitted a clear distinction between benign and tumor tissue, as p63 is only expressed in the basal cell layer of benign prostate glands. Whereas HSP60 immunoreactivity was weak or absent in the epithelium of benign tissue, tumor cells in GS6 as well as GS8+ cancers demonstrated intermediate and strong reactions to anti-HSP60 antibodies, respectively. The nuclear protein lamin A showed medium to strong expression in benign epithelium and stromal cells. In the epithelial cell compartment immunoreactivity was almost exclusively localized to the basal cell layer whereas secretory
research articles
Lamin A as a Discriminatory Biomarker
Figure 3. Clustering of differentially expressed proteins. Sample preparation and data processing as described in Materials and Methods. Data were analyzed using hierarchical clustering (red, overexpressed proteins; green, down-regulated proteins; white, missing values).
epithelial cells showed no or very weak staining. No lamin A protein staining was seen in the majority of GS6 cancers (Figure 5B); just some of them showed a few positive tumor cells (