The Prostate
Urinary Prostate Protein Glycosylation Prof|ling as a Diagnostic Biomarker for Prostate Cancer Tijl Vermassen,1 Charles Van Praet,2 Nicolaas Lumen,2 Karel Decaestecker,2 Dieter Vanderschaeghe,3 Nico Callewaert,3 Geert Villeirs,4 Piet Hoebeke,2 Simon Van Belle,1 Sylvie Rottey,1 and Joris Delanghe5* 1
Department of Medical Oncology,Ghent University Hospital,Ghent, Belgium 2 Department of Urology,Ghent University Hospital,Ghent, Belgium 3 Unit for Medical Biotechnology, Inflammation Research Center,VIB ^ Ghent University,Ghent, Belgium 4 Department of Radiologyand Nuclear Medicine,Ghent University Hospital,Ghent, Belgium 5 Department of Clinical Chemistry,Ghent University Hospital,Ghent, Belgium
BACKGROUND. Serum prostate-specific antigen (sPSA) measurement is widely used as opportunistic screening tool for prostate cancer (PCa). sPSA suffers from considerable sensitivity and specificity problems, particularly in the diagnostic gray zone (sPSA 4–10 mg/L). Furthermore, sPSA is not able to discriminate between poorly-, moderately-, and welldifferentiated PCa. We investigated prostatic protein glycosylation profiles as a potential PCa biomarker. METHODS. Differences in total urine N-glycosylation profile of prostatic proteins were determined between healthy volunteers (n ¼ 54), patients with benign prostate hyperplasia (BPH; n ¼ 93) and newly diagnosed PCa patients (n ¼ 74). Variations in N-glycosylation profile and prostate volume were combined into one urinary glycoprofile marker (UGM). Additionally, differences in N-glycosylation were identified between Gleason 7. RESULTS. The UGM was able to discriminate BPH from PCa, overall and in the diagnostic gray zone (P < 0.001). The UGM showed comparable diagnostic accuracy to sPSA, but gave an additive diagnostic value to sPSA (P < 0.001). In the diagnostic gray zone the UGM performed significantly better than sPSA (P < 0.001). A significant difference was found in corefucosylation of biantennary structures and overall core-fucosylation of multiantennary structures between Gleason < 7 and Gleason > 7 (P ¼ 0.010 and P ¼ 0.020, respectively) and between Gleason ¼ 7 and Gleason > 7 (P ¼ 0.011 and P ¼ 0.025, respectively). CONCLUSIONS. The UGM shows high potential as PCa biomarker, particularly in the diagnostic gray zone. Further research is needed to validate these findings. Prostate # 2014 Wiley Periodicals, Inc.
KEY WORDS: benign prostate hyperplasia; diagnostic marker; Gleason score; prostate cancer; urinary asparagine-linked glycosylation
INTRODUCTION Prostate cancer (PCa) is the second most common cancer diagnosed worldwide and the sixth most common cause of cancer-related death in males [1]. In prostate disease, such as PCa, the basement membrane can be disrupted and prostate specific antigen (PSA) can access the peripheral circulation [2]. In opportunistic screening for PCa, digital rectal examination (DRE) and measurement of serum PSA (sPSA) are the most frequently used tools [3]. However, this ß 2014 Wiley Periodicals, Inc.
Grant sponsor: Clinical Research Fund of the Ghent University Hospital; Grant sponsor: Scientific Research-Flanders (FWO). Conflict of interest: The authors have declared no conflict of interest.
Correspondence to: Prof. Dr. Joris Delanghe, Department of Clinical Chemistry, Ghent University Hospital, De Pintelaan 185, B-9000 Gent, Belgium. E-mail: [email protected] Received 19 June 2014; Accepted 17 September 2014 DOI 10.1002/pros.22918 Published online in Wiley Online Library (wileyonlinelibrary.com).
2
Vermassen et al.
parameter is associated with considerable sensitivity and specificity problems, especially in the “diagnostic gray zone” (sPSA between 4 and 10 mg/L) [2,4,5]. Screening for PCa using sPSA is associated with the following diagnostic problems: a substantial amount of false-positive results (and unnecessary biopsies and anxiety among patients), because prostatitis and benign prostate hyperplasia (BPH) are also associated with elevated sPSA [3,6]; a substantial amount of false-negative results (and thus missed PCa) as there is no safe cut-off value under which no PCa exists [4,7]; no differentiation between indolent and clinically significant PCa with risk of overdiagnosis and overtreatment [3,4,6,8,9]. In cancer, the glycosylation machinery such as glycosyltransferases/glycosidases are disturbed which can result in cancer-specific glycosylation changes [10]. As aberrant glycosylation of proteins is a fundamental characteristic of tumorogenesis and aggressive clinical behavior, the role of this post-translational modification has been studied in various cancer types [11,12]. In PCa, efforts have been made to improve the diagnostic potential of PSA by focusing on cancerspecific PSA glycoforms [5,13]. PSA N-glycans from normal and tumor sources are both of the complex type that differ in their content of N-acetyl neuraminic acid and fucose, thus giving rise to distinct carbohydrate epitopes [14–16]. These might serve as a new subset of PCa biomarkers [17]. New techniques have been developed to assess the diagnostic value of glycosylation changes in cancer. N-Glycosylation profiling of serum proteins proved an asset to discriminate fibrosis from cirrhosis and hepatocellular carcinoma [18]. Recently, we reported an adaptation of this methodology which enabled differentiating healthy volunteers (HV) from patients with BPH and PCa patients [19]. Our current objective is therefore dual. First, we want to distinguish HV and patients with BPH from PCa patients based on differences observed in the urinary N-glycosylation profile [19]. Secondly, we want to assess whether differences in N-glycosylation profile of urinary prostate proteins enable discrimination of poorly-differentiated (Gleason < 7), moderatelydifferentiated (Gleason ¼ 7), and well-differentiated (Gleason > 7) PCa. MATERIALS AND METHODS Patients From July 2012 to February 2014, a cohort of patients presenting at the Department of Urology at The Prostate
the Ghent University Hospital (n ¼ 173) were recruited and prospectively analyzed. In the cohort, prostate volume was estimated with transrectal ultrasonography (TRUS) and/or multiparametric magnetic resonance imaging (mpMRI). Furthermore, all PCa patients underwent clinical tumor staging by means of mpMRI [20,21] and additional prostate biopsy to determine the Gleason score according to the 2005 International Society of Urological Pathology consensus [22]. All PCa patients (n ¼ 74) were classified based on elevated PSA density (PSA/prostate volume > 0.15 ng/ml2), abnormalities on DRE, abnormalities on mpMRI, and histological confirmation after prostate biopsy. Among them, 33 patients presented with an sPSA in the diagnostic gray zone. On prostate biopsy, 19 patients had a Gleason score 6, 36 patients Gleason score 7, and 19 patients Gleason score 8 or higher. BPH patients (n ¼ 99) had no clinical suspicion of PCa: PSA density 0.05). Diagnostic Accuracy of sPSA,UGM, and Combination of Both to Distinguish BPHFrom PCa: Overall Population and Diagnostic Gray Zone Overall AUC of the UGM was 0.71 (sensitivity: 46%, specificity: 84%) compared to 0.75 for sPSA
5
(sensitivity: 45%, specificity: 89%; Fig. 2a, P ¼ 0.504). In the diagnostic gray zone the UGM demonstrated an AUC of 0.81 (sensitivity: 58%, specificity: 91%) compared to 0.57 for sPSA (sensitivity: 12%, specificity: 86%; Fig. 2b). This difference was statistically significant (P ¼ 0.004). The possible use of the UGM as an additional tool to sPSA was studied using multivariate logistic regression with the following variables: sPSA, UGM and age. Overall, the UGM proved to have an additive value over sPSA (P < 0.001) while age was not retained in the model (P ¼ 0.772). The OR of the UGM (1.51 [1.27–1.80]) was slightly higher than that of sPSA (1.21 [1.10–1.32]; Table II), This combination achieved an AUC of 0.84 (sensitivity: 61%, specificity: 89%) and showed an improved accuracy (P ¼ 0.009) compared to sPSA (Fig. 2a). In the diagnostic gray zone neither sPSA (P ¼ 0.426) nor age (P ¼ 0.926) were retained in the final model next to the UGM (P < 0.001; Table II).
Differentiating Poorly-, Moderately-, and Well-Differentiated PCa Based on Changes in Urinary Glycosylation Patients were subdivided into three groups: Gleason < 7, Gleason ¼ 7, and Gleason > 7. No difference in sPSA concentration or prostate volume was noticed between these groups (P > 0.05; Fig. 3a,b). A significant decrease in core-fucosylation of biantennary structures and in overall core-fucosylation of multiantennary structures was observed between Gleason < 7 and Gleason > 7 (P ¼ 0.010 and P ¼ 0.020, respectively; Fig. 3c,d), and between Gleason ¼ 7 and Gleason > 7 (P ¼ 0.011 and P ¼ 0.025, respectively; Fig. 3c,d). No significant differences were found between Gleason < 7 and Gleason ¼ 7. MRI proved able to differentiate Gleason < 7 from Gleason ¼ 7 and Gleason < 7 from Gleason > 7 (P ¼ 0.020 and P ¼ 0.007, respectively).
DISCUSSION Our research demonstrated the diagnostic value of N-glycosylation profiling from urinary prostate proteins, acquired through a non-invasive method (squeezing of the prostate gland), as a biomarker for PCa. First, we determined serum and urinary biochemical markers. None of these markers, except for sPSA, were able to differentiate BPH from PCa. Urinary albumin concentration was increased in PCa patients, but was within normal biochemical parameters and is most likely not a result of PCa. Secondly, we investigated the clinical utility of urinary N-glycosylation analysis. Differences in overall The Prostate
6
Vermassen et al.
Fig. 2. ROC curves analyses for the detection of PCa compared to BPH. Diagonal segments are produced by ties. a: Overall ROC curve analysis demonstrated AUC of 0.76, 0.71, and 0.84 for sPSA,UGM, and the combination of sPSA and UGM, respectively. Differences in AUC were significant between the combination of markers and UGM and sPSA (P < 0.001and P ¼ 0.009, respectively) but not between sPSA and UGM as isolated test (P ¼ 0.504). b: In the diagnostic gray zone with sPSA concentration between 4 and10 mg/L AUC of UGM (0.81) proved statisticallybetter (P ¼ 0.004) compared to sPSA (0.57).
core-a-1,6-fucosylation and total amount of triantennary structures between BPH and PCa gained in significance compared to our previously reported data [19], due to increased numbers of participants in the patient cohort. The UGM, comprised of the differences found in N-glycosylation and prostate volume, proved able to distinguish patients with BPH from PCa patients. Combining the UGM with sPSA improved the diagnostic performance for PCa detection compared to sPSA as isolated test, even though the UGM alone did not perform better than sPSA. In the diagnostic gray zone, the UGM was significantly different between BPH and PCa. Here, this test outperformed sPSA, which was useless as a diagnostic tool for PCa.
Next, we determined if well-differentiated PCa could be distinguished from moderately- and poorlydifferentiated PCa. No differences were observed in biochemical markers or prostate volume. However, when analyzing changes in the N-glycan profile, a difference was observed in core-fucosylation of biantennary and multiantennary structures between Gleason < 7 and Gleason > 7 and between Gleason ¼ 7 and Gleason > 7. A larger patient cohort of PCa patients is needed to verify whether these markers would be able to distinguish Gleason < 7 from Gleason ¼ 7. mpMRI of the prostate was able to differentiate Gleason < 7 from Gleason ¼ 7 and Gleason < 7 from Gleason > 7. However mpMRI is an expensive and timeconsuming evaluation, requiring a specifically trained
TABLE II. Multivariate Logistic Regression Diagnostic gray zone (sPSA between 4 and 10 mg/L)
Overall Variables Age sPSA UGM
The Prostate
OR (95% CI)
P
OR (95% CI)
P
1.01 (0.96, 1.06) 1.21 (1.10, 1.32) 1.51 (1.27, 1.80)
n.s.
Urinary Prostate Protein Glycosylation Prof|ling as a Diagnostic Biomarker for Prostate Cancer Tijl Vermassen,1 Charles Van Praet,2 Nicolaas Lumen,2 Karel Decaestecker,2 Dieter Vanderschaeghe,3 Nico Callewaert,3 Geert Villeirs,4 Piet Hoebeke,2 Simon Van Belle,1 Sylvie Rottey,1 and Joris Delanghe5* 1
Department of Medical Oncology,Ghent University Hospital,Ghent, Belgium 2 Department of Urology,Ghent University Hospital,Ghent, Belgium 3 Unit for Medical Biotechnology, Inflammation Research Center,VIB ^ Ghent University,Ghent, Belgium 4 Department of Radiologyand Nuclear Medicine,Ghent University Hospital,Ghent, Belgium 5 Department of Clinical Chemistry,Ghent University Hospital,Ghent, Belgium
BACKGROUND. Serum prostate-specific antigen (sPSA) measurement is widely used as opportunistic screening tool for prostate cancer (PCa). sPSA suffers from considerable sensitivity and specificity problems, particularly in the diagnostic gray zone (sPSA 4–10 mg/L). Furthermore, sPSA is not able to discriminate between poorly-, moderately-, and welldifferentiated PCa. We investigated prostatic protein glycosylation profiles as a potential PCa biomarker. METHODS. Differences in total urine N-glycosylation profile of prostatic proteins were determined between healthy volunteers (n ¼ 54), patients with benign prostate hyperplasia (BPH; n ¼ 93) and newly diagnosed PCa patients (n ¼ 74). Variations in N-glycosylation profile and prostate volume were combined into one urinary glycoprofile marker (UGM). Additionally, differences in N-glycosylation were identified between Gleason 7. RESULTS. The UGM was able to discriminate BPH from PCa, overall and in the diagnostic gray zone (P < 0.001). The UGM showed comparable diagnostic accuracy to sPSA, but gave an additive diagnostic value to sPSA (P < 0.001). In the diagnostic gray zone the UGM performed significantly better than sPSA (P < 0.001). A significant difference was found in corefucosylation of biantennary structures and overall core-fucosylation of multiantennary structures between Gleason < 7 and Gleason > 7 (P ¼ 0.010 and P ¼ 0.020, respectively) and between Gleason ¼ 7 and Gleason > 7 (P ¼ 0.011 and P ¼ 0.025, respectively). CONCLUSIONS. The UGM shows high potential as PCa biomarker, particularly in the diagnostic gray zone. Further research is needed to validate these findings. Prostate # 2014 Wiley Periodicals, Inc.
KEY WORDS: benign prostate hyperplasia; diagnostic marker; Gleason score; prostate cancer; urinary asparagine-linked glycosylation
INTRODUCTION Prostate cancer (PCa) is the second most common cancer diagnosed worldwide and the sixth most common cause of cancer-related death in males [1]. In prostate disease, such as PCa, the basement membrane can be disrupted and prostate specific antigen (PSA) can access the peripheral circulation [2]. In opportunistic screening for PCa, digital rectal examination (DRE) and measurement of serum PSA (sPSA) are the most frequently used tools [3]. However, this ß 2014 Wiley Periodicals, Inc.
Grant sponsor: Clinical Research Fund of the Ghent University Hospital; Grant sponsor: Scientific Research-Flanders (FWO). Conflict of interest: The authors have declared no conflict of interest.
Correspondence to: Prof. Dr. Joris Delanghe, Department of Clinical Chemistry, Ghent University Hospital, De Pintelaan 185, B-9000 Gent, Belgium. E-mail: [email protected] Received 19 June 2014; Accepted 17 September 2014 DOI 10.1002/pros.22918 Published online in Wiley Online Library (wileyonlinelibrary.com).
2
Vermassen et al.
parameter is associated with considerable sensitivity and specificity problems, especially in the “diagnostic gray zone” (sPSA between 4 and 10 mg/L) [2,4,5]. Screening for PCa using sPSA is associated with the following diagnostic problems: a substantial amount of false-positive results (and unnecessary biopsies and anxiety among patients), because prostatitis and benign prostate hyperplasia (BPH) are also associated with elevated sPSA [3,6]; a substantial amount of false-negative results (and thus missed PCa) as there is no safe cut-off value under which no PCa exists [4,7]; no differentiation between indolent and clinically significant PCa with risk of overdiagnosis and overtreatment [3,4,6,8,9]. In cancer, the glycosylation machinery such as glycosyltransferases/glycosidases are disturbed which can result in cancer-specific glycosylation changes [10]. As aberrant glycosylation of proteins is a fundamental characteristic of tumorogenesis and aggressive clinical behavior, the role of this post-translational modification has been studied in various cancer types [11,12]. In PCa, efforts have been made to improve the diagnostic potential of PSA by focusing on cancerspecific PSA glycoforms [5,13]. PSA N-glycans from normal and tumor sources are both of the complex type that differ in their content of N-acetyl neuraminic acid and fucose, thus giving rise to distinct carbohydrate epitopes [14–16]. These might serve as a new subset of PCa biomarkers [17]. New techniques have been developed to assess the diagnostic value of glycosylation changes in cancer. N-Glycosylation profiling of serum proteins proved an asset to discriminate fibrosis from cirrhosis and hepatocellular carcinoma [18]. Recently, we reported an adaptation of this methodology which enabled differentiating healthy volunteers (HV) from patients with BPH and PCa patients [19]. Our current objective is therefore dual. First, we want to distinguish HV and patients with BPH from PCa patients based on differences observed in the urinary N-glycosylation profile [19]. Secondly, we want to assess whether differences in N-glycosylation profile of urinary prostate proteins enable discrimination of poorly-differentiated (Gleason < 7), moderatelydifferentiated (Gleason ¼ 7), and well-differentiated (Gleason > 7) PCa. MATERIALS AND METHODS Patients From July 2012 to February 2014, a cohort of patients presenting at the Department of Urology at The Prostate
the Ghent University Hospital (n ¼ 173) were recruited and prospectively analyzed. In the cohort, prostate volume was estimated with transrectal ultrasonography (TRUS) and/or multiparametric magnetic resonance imaging (mpMRI). Furthermore, all PCa patients underwent clinical tumor staging by means of mpMRI [20,21] and additional prostate biopsy to determine the Gleason score according to the 2005 International Society of Urological Pathology consensus [22]. All PCa patients (n ¼ 74) were classified based on elevated PSA density (PSA/prostate volume > 0.15 ng/ml2), abnormalities on DRE, abnormalities on mpMRI, and histological confirmation after prostate biopsy. Among them, 33 patients presented with an sPSA in the diagnostic gray zone. On prostate biopsy, 19 patients had a Gleason score 6, 36 patients Gleason score 7, and 19 patients Gleason score 8 or higher. BPH patients (n ¼ 99) had no clinical suspicion of PCa: PSA density 0.05). Diagnostic Accuracy of sPSA,UGM, and Combination of Both to Distinguish BPHFrom PCa: Overall Population and Diagnostic Gray Zone Overall AUC of the UGM was 0.71 (sensitivity: 46%, specificity: 84%) compared to 0.75 for sPSA
5
(sensitivity: 45%, specificity: 89%; Fig. 2a, P ¼ 0.504). In the diagnostic gray zone the UGM demonstrated an AUC of 0.81 (sensitivity: 58%, specificity: 91%) compared to 0.57 for sPSA (sensitivity: 12%, specificity: 86%; Fig. 2b). This difference was statistically significant (P ¼ 0.004). The possible use of the UGM as an additional tool to sPSA was studied using multivariate logistic regression with the following variables: sPSA, UGM and age. Overall, the UGM proved to have an additive value over sPSA (P < 0.001) while age was not retained in the model (P ¼ 0.772). The OR of the UGM (1.51 [1.27–1.80]) was slightly higher than that of sPSA (1.21 [1.10–1.32]; Table II), This combination achieved an AUC of 0.84 (sensitivity: 61%, specificity: 89%) and showed an improved accuracy (P ¼ 0.009) compared to sPSA (Fig. 2a). In the diagnostic gray zone neither sPSA (P ¼ 0.426) nor age (P ¼ 0.926) were retained in the final model next to the UGM (P < 0.001; Table II).
Differentiating Poorly-, Moderately-, and Well-Differentiated PCa Based on Changes in Urinary Glycosylation Patients were subdivided into three groups: Gleason < 7, Gleason ¼ 7, and Gleason > 7. No difference in sPSA concentration or prostate volume was noticed between these groups (P > 0.05; Fig. 3a,b). A significant decrease in core-fucosylation of biantennary structures and in overall core-fucosylation of multiantennary structures was observed between Gleason < 7 and Gleason > 7 (P ¼ 0.010 and P ¼ 0.020, respectively; Fig. 3c,d), and between Gleason ¼ 7 and Gleason > 7 (P ¼ 0.011 and P ¼ 0.025, respectively; Fig. 3c,d). No significant differences were found between Gleason < 7 and Gleason ¼ 7. MRI proved able to differentiate Gleason < 7 from Gleason ¼ 7 and Gleason < 7 from Gleason > 7 (P ¼ 0.020 and P ¼ 0.007, respectively).
DISCUSSION Our research demonstrated the diagnostic value of N-glycosylation profiling from urinary prostate proteins, acquired through a non-invasive method (squeezing of the prostate gland), as a biomarker for PCa. First, we determined serum and urinary biochemical markers. None of these markers, except for sPSA, were able to differentiate BPH from PCa. Urinary albumin concentration was increased in PCa patients, but was within normal biochemical parameters and is most likely not a result of PCa. Secondly, we investigated the clinical utility of urinary N-glycosylation analysis. Differences in overall The Prostate
6
Vermassen et al.
Fig. 2. ROC curves analyses for the detection of PCa compared to BPH. Diagonal segments are produced by ties. a: Overall ROC curve analysis demonstrated AUC of 0.76, 0.71, and 0.84 for sPSA,UGM, and the combination of sPSA and UGM, respectively. Differences in AUC were significant between the combination of markers and UGM and sPSA (P < 0.001and P ¼ 0.009, respectively) but not between sPSA and UGM as isolated test (P ¼ 0.504). b: In the diagnostic gray zone with sPSA concentration between 4 and10 mg/L AUC of UGM (0.81) proved statisticallybetter (P ¼ 0.004) compared to sPSA (0.57).
core-a-1,6-fucosylation and total amount of triantennary structures between BPH and PCa gained in significance compared to our previously reported data [19], due to increased numbers of participants in the patient cohort. The UGM, comprised of the differences found in N-glycosylation and prostate volume, proved able to distinguish patients with BPH from PCa patients. Combining the UGM with sPSA improved the diagnostic performance for PCa detection compared to sPSA as isolated test, even though the UGM alone did not perform better than sPSA. In the diagnostic gray zone, the UGM was significantly different between BPH and PCa. Here, this test outperformed sPSA, which was useless as a diagnostic tool for PCa.
Next, we determined if well-differentiated PCa could be distinguished from moderately- and poorlydifferentiated PCa. No differences were observed in biochemical markers or prostate volume. However, when analyzing changes in the N-glycan profile, a difference was observed in core-fucosylation of biantennary and multiantennary structures between Gleason < 7 and Gleason > 7 and between Gleason ¼ 7 and Gleason > 7. A larger patient cohort of PCa patients is needed to verify whether these markers would be able to distinguish Gleason < 7 from Gleason ¼ 7. mpMRI of the prostate was able to differentiate Gleason < 7 from Gleason ¼ 7 and Gleason < 7 from Gleason > 7. However mpMRI is an expensive and timeconsuming evaluation, requiring a specifically trained
TABLE II. Multivariate Logistic Regression Diagnostic gray zone (sPSA between 4 and 10 mg/L)
Overall Variables Age sPSA UGM
The Prostate
OR (95% CI)
P
OR (95% CI)
P
1.01 (0.96, 1.06) 1.21 (1.10, 1.32) 1.51 (1.27, 1.80)
n.s.
