Four potential biomarkers as prognostic factors in stage III serous ovarian adenocarcinomas

Abstract

The mortality rate for patients with ovarian carcinomas is high and the available prognostic factors are insufficient. The use of biomarkers may contribute to better prediction and survival for these patients. We aimed to study the gene and protein expressions for 7 potential biomarkers, to determine if it is possible to use them as prognostic factors. Genes selected from our previous microarray analysis (2006), CLU, ITGB3, TACC1, MUC5B, CAPG, PRAME and TROAP, were analyzed in 19 of the tumors with quantitative real-time polymerase chain reaction (QPCR). We found that CLU and ITGB3 were more expressed in tumors from survivors and PRAME and CAPG were more expressed in tumors from deceased patients. None of the other 3 genes were significantly differently expressed. The protein expressions of CLU, ITGB3, PRAME and CAPG were analyzed in 43 of the tumors with western blot for semiquantitative analysis. We established that the mRNA and protein expressions correlated and that all 4 proteins were significantly differently expressed. Further, immunohistochemistry (IHC) was used to localize the expression of the proteins in the tumor samples. According to our results, the 4 biomarkers CLU, ITGB3, PRAME and CAPG may be used as prognostic factors for patients with stage III serous ovarian adenocarcinomas. © 2008 Wiley-Liss, Inc.

Ovarian carcinoma is ranked 10th in cancer incidence in Swedish women, yet it is the 5th most common cause of cancer death in that population.1, 2 Over 70% of women with ovarian carcinoma are diagnosed with advanced disease, stage III or IV.3 It is difficult to cure patients in advanced stages of the disease and the mortality rate is high: only 30% of patients survive more than 5 years after diagnosis.4

The most frequent type is epithelial ovarian adenocarcinoma, of which serous papillary adenocarcinoma accounts for 50%.4 The treatment of patients with these tumors is governed by various prognostic factors such as surgical stage, volume of residual tumor after primary surgery and histologic grade. Despite this, the clinical outcome can be difficult to predict in an individual patient with advanced disease. Some patients will respond to surgery followed by chemotherapy, whereas others with the same prognosis and treatment will relapse or their tumors continue to grow. If these different patient groups could be identified before treatment, then alternative strategies could be used instead of standard therapy to optimize the treatment of each patient.

The use of novel biomarkers may contribute to better survival for patients with ovarian tumors. In our previous microarray analysis of 54 stage III serous ovarian adenocarcinomas, we detected a gene expression profile representing tumors from a group of survivors.5 However, microarray experiments accumulate a great quantity of data, and further evaluation of the findings is needed to draw conclusions about their applicability as prognostic markers. Microarray analyses of ovarian carcinomas concerning survival have successfully created gene profiles that can distinguish between short- and long-term survivors or predict the response to platinum-based chemothreapy.6, 7 Because mRNA is quickly degraded if samples are not prepared correctly, proteins, which are more stable, might be more useful as biomarkers. Differences in protein expressions may also be relevant to the tumors' biological properties. Moreover, there are several detection alternatives for proteins. The use of labeled antibodies in western blot, immunohistochemistry (IHC) and flow cytometry might be preferable for detection of biomarkers. In addition, proteins may be detected in ascites or blood samples, as CA-125 levels are monitored in ovarian cancer patients.

In this study, we aimed to verify the differences in gene expressions identified with microarray and to establish if the differences are detectable at the protein level. Seven cancer-related genes were selected from our microarray analysis: CAPG, CLU, ITGB3, MUC5B, PRAME, TACC1 and TROAP. The gene expressions were quantified by quantitative real-time polymerase chain reaction (QPCR). The corresponding proteins of differently expressed genes (CLU, ITGB3, PRAME and CAPG) were analyzed with western blot for semiquantitative analysis to identify profiles and/or single molecular markers with prognostic relevance.

cCLU, cytoplasmic CLU; FIGO, International Federation of Gynaecology and Obstetrics; IHC, immunohistochemistry; nCLU, nuclear CLU; QPCR, quantitative real-time polymerase chain reaction; sCLU, secreted CLU; TBS, Tris buffered saline; TBS-S, TBS containing 0.1% saponin.

Material and methods

Patient and tumor material

In this study, 43 of the 54 previously studied stage III serous papillary adenocarcinomas of the ovary were used (Table I).5 The number of tumors is reduced because of lack of tumor material in 9 samples. All samples were from the peripheral part of primary tumors. Patients who survived 5 years or more after the initial diagnosis were considered as survivors and all deceased patients in the study succumbed to cancer. Surgical staging of the tumors was performed according to International Federation of Gynaecology and Obstetrics (FIGO) standards and patients with no macroscopic residual tumor were classified as radically operated. The tumors were removed at primary surgery and secured for pathology examination, RNA and protein extraction. After surgery, patients were treated with a combination of farmorubicine, carboplatin and cyclophosphamide. The tumors were collected from patients diagnosed between 1993 and 2000 at Sahlgrenska University Hospital, Gothenburg, Sweden, and the study was approved by the local ethics committee. A pathologist reviewed all tumors according to the treatment protocol for gynecological malignancies in western Sweden. Specimen imprints for cytological evaluation were performed to verify the presence of tumor cell content and only tumors containing at least 50% tumor cells were included.

Table I. Summary of Clinicopathologic Characteristics of the Patients

	Tumours (from survivors/deceased patients)
Total tumours	43 (16/27)
Tumours used in QPCR analysis	19 (9/10)
Mean age (years)	58 (52/62) (range 35–82)
FIGO staging
IIIa	6 (5/1)
IIIb	15 (8/7)
IIIc	22 (3/19)
Surgery
Radically operated	10 (8/2)
Residual tumour	33 (8/25)
Differentiation
Well	6 (5/1)
Moderate	15 (8/7)
Poor	22 (3/19)

Gene selection

In our earlier study, 54 tumors were analyzed with microarray.5 These data were used to search for genes to investigate in the present study. The data generated a list with 11 genes that were expressed differently in tumors from survivors and from deceased patients with false discovery rate (FDR) adjusted p-values ≤ 0.1 and at least a 2-fold change between the groups. We used a combination of this list and the available information about the genes' function and cancer connection to select the 7 genes analyzed in this study (Table II). Four of the 7 genes, CLU, ITGB3, TACC1 and MUC5B were more expressed in tumors from survivors, and 3 of the 7 genes, CAPG, PRAME and TROAP were more expressed in tumors from deceased patients.

Table II. Genes Differently Expressed in Tumours from Survivors Compared with Tumours from Deceased Patients with at Least 2-Fold Change and Adjusted p-VALUE < 0.1

Symbol	Name	Fold change	p-value
TACC1	Transforming acidic coiled-coil containing protein 1	2.86	0.02
MUC5B	Mucin 5B	2.73	0.06
ITGB3	Integrin beta 3	2.67	0.10
CLU	Clusterin	2.65	0.06
CAPG	Capping protein (actin filament) gelsolin-like	−2.11	0.10
TROAP	Trophinin associated protein (tastin)	−2.15	0.10
PRAME	Preferentially expressed antigen in melanoma	−2.56	0.06

Quantitative real-time polymerase chain reaction

QPCR was performed on 19 samples as described by Partheen et al.5 with minor modifications. Briefly, total RNA was isolated from frozen tumors by homogenization with TRIzol Reagent (Invitrogen, Carlsbad, CA) and then extracted with RNeasy mini kit (Qiagen, Valencia, CA). The 19 samples with highest RNA quality from the previous microarray analysis, 9 from survivors and 10 from deceased patients, were selected for evaluation of the 7 genes. From each tumor sample, 0.5 μg of total RNA was reverse transcribed in duplicate. Each cDNA sample was analyzed once by real-time PCR, giving 2 data points for each tumor sample. The cDNA were detected with SYBR Green I. Primer sequences for the target genes are presented in Table III. Two reference genes were used for normalization, GAPDH and β-actin, from the Human Endogenous Control Gene Panel (TATAA Biocenter, http://www.tataa.com). The efficiency of each QPCR assay was estimated from the slope of a standard curve generated from the serial dilution of purified PCR products. The assays for CLU and ITGB3 showed PCR efficiencies close to 80% and the remaining genes 90%. These values were used for subsequent calculations. For each assay, the average Ct value for each tumor sample was converted to relative copy numbers. The data were then normalized by the geometric average of the 2 reference genes.8

Table III. Primer Sequences Used in QPCR

Gene	Forward and reverse primer
CAPG	5′-agcatttcacaagacctcca
	5′-caccacaccaggcgaaga
CLU	5′-cggatgaaggaccagtgtg
	5′-agcaaggaggaggtgttgag
ITGB3	5′-cattactctgcctccactacca
	5′-aacggattttcccataagca
TACC1	5′-aactccccacccctctctt
	5′-cttcttcaccttacagccactc
MUC5B	5′-tgccccttgttctgtgactt
	5′-acgcacttcatctggtcctc
PRAME	5′-tatcgcccagttcacctctc
	5′-gggacttacatcggtcagca
TROAP	5′-ttcctttcacccttccactc
	5′-atggctgtccctggtatgt

Western blot

Western blot was performed on 43 tumor samples, each in duplicate. Total protein extracts were prepared in RIPA lysis buffer containing sodium orthovanadate, (100 mM, 10 μl/ml), phenylmethanesulfonyl fluoride (1 μg/ml) and aprotinin (50 kallikrein inhibitory units per ml) (Santa Cruz Biotechnology, Santa Cruz, CA). The frozen samples were homogenized with RIPA in a Mikro-Dismembrator S and by sonication, followed by centrifugation (10,000g, 20 min, 4°C). Supernatants were stored at −80°C until analysis. The protein concentrations were determined with the BCA™ protein assay kit, according to the manufacturer's instructions (Pierce, Rockford, IL). The samples were diluted in SDS sample buffer with and without 10% 2-mercaptoethanol and heated at 97°C for 5 min. The unreduced samples (without 2-mercaptoethanol) were used to detect ITGB3. Twenty micrograms of total protein was loaded into each lane on a 10% Criterion™ precast gel (BIO-RAD laboratories, Hercules, CA) with MagicMark™ XP western protein standard (Invitrogen) used as weight marker. The proteins were transferred to nitrocellulose membranes (BIO-RAD) and blocked overnight at 4°C in 5% nonfat milk in 10 mM Tris buffered saline (TBS).

The membranes were incubated in TBS containing 0.05%Tween 20 with the following primary antibodies: chicken polyclonal to CAPG (1:3,000, ab14235, Abcam, Cambridge, UK), rabbit polyclonal to PRAME (1:1,000, ab32185, Abcam), mouse monoclonal to CLU (1:15,000, clone 41D, Upstate Biotechnology, Lake Placid, NY), ITGB3 (1:1,000, MAB1974, Chemicon, Temecula, CA) and GAPDH as loading control (1:15,000, ab8245, Abcam). Proteins were visualized by chemiluminescence, using horseradish peroxidase-linked (HRP) secondary antibodies; anti-chicken (1:3,000, GAYFC-HRP, Genway Biotech, San Diego, CA), goat anti-rabbit (1:5,000, ab6721, Abcam) and anti-mouse (1:5,000, sc-2005, Santa Cruz Biotechnology), enhanced by ECL™ Western Blotting Detection reagents (Amersham, Buckinghamshire, UK). The membranes were exposed to Amersham Hyperfilm™ ECL (Amersham). The optical density from each band was measured using the software package Quantity One (BIO-RAD) and used for semiquantitative analysis of the proteins. An internal reference sample, same on each blot, was used as a standard for quantification and was given the value 1.9

Immunohistochemistry

Fresh-frozen tissues from 43 tumors were cryosectioned (5 μm), fixed in cold acetone for 1 min and then dried at room temperature and stored at −80°C. The staining procedure was accomplished using DakoCytomation EnVision⁺ HRP according to the manufacturer's instructions (Dako, Carpinteria, CA). TBS (100 mM) containing 0.1% saponin (TBS-S) was used as buffer solution. Prior to incubation with CLU antibody, slides were blocked with TBS-S containing 3% BSA. Slides were incubated for 30 min with primary antibody mouse monoclonal to CLU (1:3,000, clone 41D, Upstate Biotechnology), mouse monoclonal to ITGB3 (1:100, ab7167, Abcam), rabbit polyclonal to PRAME (1:50, ab32185, Abcam) or with chicken polyclonal to CAPG (1:3,000, ab14235, Abcam), diluted in TBS-S containing 1% BSA. The CLU, PRAME and CAPG antibody used was the same as in the immunoblotting assay, but the 2 ITGB3 antibodies used were different, because none of them worked in both applications. Secondary antibody anti-chicken (1:100, GAYFC-HRP, Genway Biotech) was used to detect CAPG. The incubation buffer without primary antibody was used as negative control and staining in endothelial cells was used as internal positive control for ITGB3. The sites of peroxidase binding were detected with 3,3′-diaminobenzidine and cells were counterstained with hematoxylin.

Statistical analysis

Statistical analysis of the data was performed using SPSS for Windows (v 12.0.1). The relation between expressions measured with the different methods microarray, QPCR and western blot was evaluated with bivariate correlation using Spearman correlation coefficient. Statistical differences in mRNA and protein expressions between tumors from survivors and tumors from deceased patients were evaluated using the Mann-Whitney U-test. Further, Kaplan-Meier survival curves were used to show differences in clinical outcome between patients with tumors that expressed high alternatively low levels of protein. A log-rank test was used to compare the curves. A value of p < 0.05 was considered to be significant.

A logistic regression model was setup to investigate if the 4 protein expressions could contribute to the prediction of the clinical outcome. Because of the skewed expression distributions, the logarithms of protein expressions were used as explanatory variables in the model. Forward and backward stepwise selections were used to pick out the best combination of explanatory variables among the standard prognostic factors and the 4 proteins and their 2-ways interactions.

Results

Four genes were detected as differently expressed

Seven genes, CLU, ITGB3, TACC1, MUC5B, CAPG, PRAME and TROAP, were analyzed with QPCR in the 19 tumors with the highest RNA quality from our previous microarray study. The relative copy number in 1 sample (47D) for MUC5B was extremely low and therefore excluded. The reliability of our expression data was confirmed by correlation analysis with the expression levels previously detected with microarray, where all genes correlated well (p < 0.01). When tumors from survivors were compared to tumors from deceased patients, 3 genes were significantly differently expressed: CLU (p = 0.022) and ITGB3 (p = 0.003) were more expressed in tumors from survivors and PRAME (p = 0.009), was more expressed in tumors from deceased patients (Fig. 1). In addition, CAPG (p = 0.094) was more expressed in tumors from deceased patients and also included in further analysis. The other genes were not differently expressed: TACC1 (p = 0.22), MUC5B (p = 0.27) and TROAP (p = 0.69).

Figure 1. Relative mRNA copy numbers as log₂ values for the differently expressed genes (a) CLU, (b) ITGB3, (c) PRAME and (d) CAPG. Each plot shows the medians (centre lines), interquartile ranges (boxes), largest and smallest values (whiskers) that are not outliers (circles), or extreme values (stars) within a category.

Differences were detected for all 4 protein expressions

The corresponding proteins to the 4 genes detected as differently expressed with QPCR were examined with western blot for semiquantitative measurements of each protein in 43 samples (Fig. 2). The differences in expression in tumors from survivors and from deceased patients were also detected at the protein level and were significant for all 4 proteins; CLU (p = 0.002), ITGB3 (p = 0.025), PRAME (p = 0.008) and CAPG (p = 0.015). The Kaplan-Meier survival curves for high versus low expression for each protein are shown in Figure 3. There were significant differences in survival for tumors with high versus low expression of each protein. We also established a correlation between the expression levels detected with western blot and QPCR in 2 of 4 cases, CAPG (p < 0.001), and CLU (p = 0.01) but not for ITGB3 (p =0.06) and PRAME (p = 0.13). The antibody used to detect CLU detects all isoforms of the protein (nCLU, cCLU and sCLU, see discussion).10 The nCLU signal was not evaluated, because it was overshadowed by the sCLU signal for a few samples. Instead, IHC analysis was performed to detect any nuclear localization of the protein.

Figure 2. Relative protein expressions and a representative immunoblot for (a) CLU, (b) ITGB3, (c) PRAME and (d) CAPG. GAPDH was used as loading control and tumors from survivors (S) and deceased (D) patients are indicated. Each plot shows the medians (centre lines), interquartile ranges (boxes), largest and smallest values (whiskers) that are not outliers (circles), or extreme values (stars) within a category.

Figure 3. Kaplan-Meier survival curves for (a) CLU, (b) ITGB3, (c) PRAME and (d) CAPG.

The prognostic factors available today and the protein expressions were further subjected to logistic regression analysis to determine if a classification of the tumors could be identified. If the tumors were classified with our 4 protein expressions, 88.4% of the tumors were classified correctly and 2/27 tumors from deceased patients were incorrectly classified as “survivors.” A classification with the standard prognostic factors age, stage, grade and residual tumor after surgery generated the same results. A combination of the 4 biomarkers and the standard prognostic factors classified 90.7% of the tumors correctly, and 1/27 tumors from deceased patients were incorrectly classified as “survivor,” which is a minor improvement of the results. A combination of only CLU and PRAME also generated a sufficient result; 83.7% of the tumors were classified correctly.

Localization of protein with Immunohistochemistry

IHC staining was performed to detect if CLU was expressed as nCLU, cCLU or sCLU (Figs. 4a–4d). The staining revealed that CLU was not expressed in the nucleus in any tumor, which was in agreement with our western blot results. Positive staining was detected in the cytoplasm and outside the cell membrane (Figs. 4a and 4b). Both tumor epithelial cells and stroma cells were stained at different levels in the samples and none were completely negative. Interestingly, 14 tumors were only positively stained in tumor stroma cells (Figs. 4c and 4d) and 12 of these tumors were from deceased patients.

Figure 4. Positive immunohistochemistry staining of CLU (a–d), ITGB3 (e–h), CAPG (i,j) and PRAME (k,l) in serous ovarian adenocarcinoma. Positive staining of CLU in tumor epithelial (e→) and stroma (s→) cells from a survivor in (a), with a magnified view of the box in (b). Positive staining of CLU in tumor stroma cells from a deceased patient in (c), with a magnified view of the box in (d). Positive staining of ITGB3 in tumor cells from a survivor in (e) with a magnified view of the box in (f). Positive staining of ITGB3 in tumor stroma cells (g). Negative staining of ITGB3 in tumor cells from a deceased patient (h), with positive staining in endothelial cells (→) used as internal positive control. Positive staining of CAPG in tumor cells from a deceased patient in (i) with a magnified view of the box in (j). Positive staining of PRAME in tumor cells from a deceased patient in (k) with a magnified view of the box in (l).

ITGB3 is expressed on tumor cells but also on endothelial cells, monocytes, platelets and megakaryocytes. IHC was used to ensure that tumor cells expressed ITGB3. Positive tumor staining was detected in 23 of 43 samples (Figs. 4e and 4f). The staining was detected in epithelial tumor membrane and cytosol and on endothelial cells. In 5 samples, positive staining was detected in tumor stroma (Fig. 4g) and 2 of those were negative in epithelial cells. Four of the 20 negative tumor samples were from survivors and all showed positive staining in endothelial cells (Fig. 4h), which was used as internal positive control.

CAPG positive staining was detected in the cytoplasm of tumor epithelial cells and stroma cells at different levels in 39 samples (Figs. 4i and 4j). The staining showed that CAPG was not expressed in the nucleus in any tumor. However, the lack of positive control for nuclear staining has to be considered in the interpretation of the result. Positive staining of PRAME was detected in 33 tumors (Figs. 4k and 4l). The staining was detected in the cytoplasm of tumor epithelial cells and stroma cells in clusters of cells in the samples. The negative controls for all antibodies where negative (data not shown).

Discussion

In this study, we aimed to analyze a selection of genes and proteins and the possibility to use them as biomarkers. We had previously analyzed tumors with microarray and verified the data with QPCR.5 In that study, we found that the QPCR analysis required higher mRNA quality to get reliable results. Therefore, a selection of 19 tumors with the highest RNA quality was used in this study to validate differences in gene expressions between tumors from survivors and tumors from deceased patients. Four of 7 genes, CLU, ITGB3, PRAME and CAPG were differently expressed. These genes were further evaluated. The lower number of tumors included in the QPCR analysis could explain the divergent results compared with the microarray results. Thus, the other genes should not be discarded as irrelevant in cancer progression, but further evaluation of them is required.

The translation of mRNA to proteins is not always straightforward. Posttranscriptional modifications and several other factors than mRNA concentrations control the quantity of proteins produced. The significant correlation between mRNA and protein expression for CLU, and CAPG in our study indicate that these protein expressions are related to the quantity of mRNA. Further, our protein analysis showed that it was possible to detect variation in protein expressions. All 4 proteins were significantly differently expressed between tumors from survivors and from deceased patients. The differences in protein expressions, and not only gene expressions, indicate that the proteins' function could be involved in tumor growth.

CLU (Clusterin), which was more expressed in tumors from survivors in our study, encodes a glycoprotein that is ubiquitously expressed at the mRNA and/or protein level in serum and many tissues such as testis, ovary, brain and liver.11–13 The mRNA expression is induced by heat shock, and it has been demonstrated that CLU has chaperone activity.14, 15 Controversially, CLU has also been shown to have proapoptotic function.16 This divergent role of the protein is probably due to different isoforms, 1 secreted form (sCLU) with a cytoplasmic precursor (cCLU) with chaperone activity involved in tumor progression and 1 nuclear form (nCLU) with proapoptotic function.14–16 CLU has been investigated in several types of cancer and reported as upregulated in various advanced tumors compared to early tumors or normal tissues, including bladder, prostate and breast cancer.17–19 On the contrary, expression of CLU has been reported as downregulated in pancreatic adenocarcinoma, prostate cancer and serous ovarian carcinoma.20–22 These inconsistent findings may depend on the different isoforms. Pucci et al. studied CLU in the different steps of colon carcinoma progression and discovered a translocation of CLU from the nucleus to the cytoplasm during tumor development.10 Further, Xie et al.23 found a significant association between overexpression of cCLU in ovarian carcinomas compared to normal ovaries, cystadenomas and borderline tumors. In our study, the protein CLU was expressed as cCLU or sCLU, which is in accordance with these previous findings. Although CLU expression increases in tumors during progression, its role as a biomarker for survival has to be clarified. The expression of cCLU has been associated with longer survival in patients with lung cancer, which is in concordance with our results.24 At first, this may seem to be in disagreement with previous findings of increased CLU expression in advanced stages. However, we investigated only advanced stage disease, and in this group CLU could be a marker for better survival, according to our results.

ITGB3 (Integrin beta 3) was also more expressed in tumors from survivors in our study. The protein associates at the cell surface with the alpha V integrin to form the αvβ3 complex on endothelial and tumor cells, monocytes, platelets and osteoclasts. In association with the alpha IIb integrin, ITGB3 forms the GPIIb/IIIa complex on platelets and megakaryocytes.25 Integrins work as receptors and participate in cell adhesion as well as cell-surface mediated signaling.26–29 The αvβ3 integrins are expressed in normal ovarian epithelium.30 Further, the expression of αvβ3 integrins has been linked to bad prognosis in breast cancer and melanoma, but the relation to ovarian tumors is less clear.31, 32 In a study by Carreiras et al.,33 ITGB3 was found in normal ovarian epithelium and highly differentiated carcinomas, but was lacking in most of the less-differentiated ovarian tumors. The expression of the β3 subunit in ovarian tumors has also been found as significantly less frequent in grade 3 than in grade 1 and 2 tumors and less expressed in peritoneal metastasis compared to primary tumors.34 Controversially, the αvβ3 integrin expression in ovarian tumor samples has been linked to ovarian tumor progression.35 Our data support the outcome in the former studies, where the protein is less expressed in advanced ovarian tumors. Further, Maubant et al.34 remark the need to describe not only the percentage of ITGB3 staining with IHC, but also the type of staining, because the function of cytoplasmic ITGB3 is less clear than the function of ITGB3 in the membrane. Our tumors were positively stained in the cytoplasm and membrane and this outcome may influence the role of ITGB3 in cancer progression.

PRAME (Preferentially expressed antigen in melanoma) was found as more expressed in tumors from deceased patients compared to tumors from survivors. To our knowledge, this is the first study of the protein PRAME in ovarian tumors. However, the mRNA expression of PRAME has been described as upregulated in malignant ovarian tumors when compared with normal ovarian tissue.36–39 The function of PRAME in normal tissue is still unknown, but it encodes an antigen recognized by autologous cytolytic T lymphocytes, and its expression is absent or low in normal adult tissue, except male germ cells.40 However, it is frequently expressed in a variety of cancers, such as melanoma and neuroblastoma and is therefore evaluated as a potential marker for advanced disease.40, 41 In a study of 31 tumors by Lancaster et al.,42 the gene expression of PRAME was upregulated in ovarian cancer samples compared to normal ovarian surface epithelium, but was not detected as differently expressed concerning survival. However, PRAME expression has been associated with unfavorable outcome for breast cancer patients and indicated as an independent prognostic factor for survival.43 These data correlate well with our results, where PRAME was more expressed in tumors from deceased patients, and promote the role of PRAME as a prognostic factor for cancer patients.

CAPG (Capping protein (actin filament) gelsolin-like), which was more expressed in tumors from deceased patients in our study, belongs to the gelsolin protein superfamily. This is a group of proteins that control actin organization and initiate actin filament growth by severing filaments, capping filament ends and nucleating actin assembly.44, 45 Nonmuscle cells use its actin filament network to change shape during movement.46 CAPG contributes to the dynamic of actin filaments by capping the barbed ends and is therefore involved in the control of actin-based motility.47 The involvement of CAPG in cancer is not precisely known, and the relation of CAPG gene and protein expressions to ovarian cancer has not been evaluated before. An elevated expression of CAPG has been detected in cancer cells compared to benign or normal cells in pancreatic carcinoma, melanoma and breast cancer cells.48–50 Further, elevated levels of CAPG trigger cellular invasion and nuclear CAPG affects the invasive phenotype, but how this relationship is regulated is not precisely known.48, 51 However, our tumors were not stained in the nucleus, which suggest that both cytoplasmic and nuclear protein can affect the tumor growth. The function of CAPG as a protein that controls migration of cells indicates that overexpression of CAPG promotes aggressive tumor progression, which correlates with our data, where CAPG was more expressed in tumors from deceased patients.

The detected differences in expressions demonstrate that each of the 4 proteins may be used as biomarkers to separate patients with diverse clinical outcomes. Moreover, a logistic regression analysis was used to create a model that combined the protein expressions. The model distinguished between tumors from survivors and tumors from deceased patients at the same level as the prognostic factors available today. The combination of our 4 biomarkers and the standard prognostic factors classified only 1 sample from a deceased patient as a “survivor.” Our model indicates that it is possible to identify a homogeneous group of survivors, which is not possible to detect with the prognostic factors used today. This subgroup of survivors with favorable clinical outcome might be the same as identified with our previous microarray cluster analysis.5 Our combination of protein expressions may be used to identify patients likely to benefit from standard therapy. These patients could be treated with a more moderate combination and the high risk patients may be exposed to additional chemotherapy and offered a more frequent follow up at an initial state.

We conclude that the genes CLU and ITGB3 were more expressed in tumors from survivors, while PRAME and CAPG were more expressed in tumors from deceased patients. Analysis of the corresponding proteins shows that it was possible to detect the differences at the protein level, as well. These differences indicate that the proteins could affect ovarian tumor growth. Our results reveal 4 potential biomarkers that may be used to biologically characterize subgroups of ovarian cancer tumors. With these novel markers, different patient groups could be identified before treatment and alternatives to standard therapy could be used to optimize the treatment of each patient.

Acknowledgements

The authors thank Linda Strömbom and Sara Andersson, TATAA Biocenter, for help with technical support of QPCR and Marita Olsson and Oscar Hammar for statistical advice.