Zinc α2-glycoprotein as a potential novel urine biomarker for the early diagnosis of prostate cancer


Ioannis Katafigiotis, Department of Urology, Athens University Medical School – LAIKO Hospital, 17 Agiou Thoma str., 11527 Athens, Greece. e-mail: katafigiotis@yahoo.com


Study Type – Diagnosis (exploratory cohort)

Level of Evidence 2b

What's known on the subject? and What does the study add?

The use of biomarkers to detect a cancer early, especially prostate cancer, is not a new idea and PSA has been proved to be the best biomarker for the early diagnosis of prostate cancer. Since the introduction and wide use of PSA various efforts have been made to find novel biomarkers in both serum and urine of individuals at high risk for prostate cancer. The best example of a biomarker detected in the urine after a vigorous digital rectal examination is PCA3, which is used mainly in the subgroup of patients with PSA 4–10 ng/mL whose prostate biopsy was repeatedly negative for prostate cancer in order to decide the performance or not of a new biopsy. Proteomics is a state of the art new biotechnology used to identify the proteome of a certain tissue meaning the whole group of proteins related to the anatomy and biochemistry of the tissue. Using proteomics can effectively and more specifically identify proteins that can be used as potential biomarkers for the early diagnosis of prostate cancer. Zinc α2-glycoprotein has been studied in the past as a protein related to cancer cachexia and it has been measured in both prostate tissue and serum in patients with prostate cancer. Zinc α2-glycoprotein has also been recently identified by proteomics in prostate tissue showing different values in patients with prostate cancer and benign prostate hyperplasia. It is the first time that zinc α2-glycoprotein has been systematically measured and studied in an easily obtained biological fluid such as urine showing a very optimistic potential both as a novel solo biomarker and as an adjunct to PSA for the early diagnosis of prostate cancer.

PSA has revolutionized the way we approximate prostate cancer diagnosis. Even though PSA is still the best biomarker for the diagnosis of prostate cancer it constitutes an organ-specific and not a disease-specific biomarker and diagnostic dilemmas are often raised concerning the performance or not of a prostate biopsy. Thus novel biomarkers are required in order to improve the diagnostic ability of PSA. Increasingly in the literature it is stated that the future of prostate cancer diagnosis could be not a single biomarker but a band of different biomarkers that as a total could give the possibility of an individual having prostate cancer. By detecting and measuring zinc α2-glycoprotein in the urine we believe that interesting conclusions can be made: first that proteomics is the way to detect with accuracy proteins that could be proved to be valuable novel biomarkers; second that zinc α2-glycoprotein detected in the urine could be used both as a solo biomarker and as an adjunct to PSA for the early diagnosis of prostate cancer.


  • • To examine the potential utility as a novel biomarker in the urine of zinc α2-glygoprotein (ZAG) for the early diagnosis of prostate cancer.


  • • The urine of 127 consecutive candidates for a transrectal ultrasound prostatic biopsy with a mean age of 65.7 ± 8.7 years and mean PSA 9.1 ± 5.3 ng/mL was collected.
  • • Western blot analysis and immunohistochemistry for ZAG were performed.
  • • Receiver operating characteristic curves and logistic regression models were used to estimate the predictive ability of ZAG and to determine the optimal sensitivity and specificity by using various cut-off values for the prediction of prostate cancer.


  • • In all, 42 patients had prostate cancer, 29 showed high grade prostatic intraepithelial neoplasia and 56 were negative.
  • • Receiver operating characteristic curve analysis showed a significant predictive ability of ZAG for prostate cancer. The area under the curve (AUC) for the prediction of prostate cancer was 0.68 (95% CI 0.59–0.78).
  • • The combination of ZAG with PSA showed a significant improvement in the predictive ability (P= 0.010), with AUC equal to 0.75 (95% CI 0.66–0.85). Separate analysis in patients with PSA levels of 4–10 ng/mL (70.1%) showed that ZAG had a discriminative power with AUC equal to 0.68.
  • • The optimal cut-off was 1.13 for ZAG, which corresponded to 6.88 times greater odds for prostate cancer.


  • • Urine detected ZAG showed promising results in the prediction of prostate cancer.
  • • Further validation is required to establish ZAG as a novel biomarker.

prostate cancer


transrectal ultrasound guided biopsy


zinc α2-glycoprotein


interquartile range


receiver operating characteristic


area under the curve


More than 240 000 new cases and more than 28 000 estimated deaths during 2012 in the USA will involve prostate cancer (PCa), which remains the most common cancer and the second cause of cancer-specific death in men [1]. Over the last 20 years, a huge rise in PCa incidence has been observed much of which can be attributed to the use of PSA testing [2]. Nevertheless, researchers have demonstrated that there is no cut-off point of PSA with sensitivity and specificity but rather a continuum of PCa risk at all values of PSA [3]. On the other hand, even when the conventional PSA threshold of 4 ng/mL is used as a transrectal ultrasound guided biopsy (TRUS-b) recommendation, 75% of men with PSA 4–10 ng/mL do not have PCa, adding unnecessary cost, anxiety and morbidity for the patient [4]. Although various PSA derivatives have been introduced, their usefulness and ability to accurately diagnose PCa remains limited [5]. As a consequence, there is still a need for novel biomarkers that could further improve our preoperative diagnostic capability.

In the present study, we evaluated the potential utility of zinc α2-glycoprotein (ZAG) in the urine of patients suspicious for PCa. ZAG is an Mr 41 000 glycoprotein secreted by a variety of normal epithelia and was recently shown to stimulate lipolysis in adipocytes, leading to the development of cachexia in animals with ZAG-producing tumours [6]. Our decision to focus on ZAG was related to the fact that recent studies have shown ZAG's association to PCa or PCa recurrence [7–9]. Moreover, following a state of the art tissue proteomic analysis conducted previously in our department, ZAG was found to be differentially expressed in the cancerous and adenomatous prostate.


This study was approved by the ethical committee of the Athens University Medical School and LAIKO Hospital (approval no. 461/22-7-09). Written informed consent was obtained. Candidate patients for a TRUS-b had an abnormal DRE or an elevated PSA (>4 ng/mL) or an abnormal PSA velocity (PSAV) of >0.75 ng/mL. Exclusion criteria consisted of patients who had already been diagnosed with PCa, patients who had been previously subjected to a TRUS-b, PSA >25 ng/mL, patients taking finasteride or dutasteride, patients who had been subjected to a surgical procedure for BPH and patients with rectal extirpation. The urine samples of 127 patients were obtained after a vigorous prostatic massage lasting 30 s and were immediately stored at –80 °C. All patients were subsequently subjected to a 10-core per prostatic lobe TRUS-b. Histological analyses of the TRUS-b specimen were done in a blinded fashion by an experienced uropathologist (K.P.).


Western blot analysis (Fig. 1)

Figure 1.

Representative image of ZAG. Western blot analysis in individual urine samples. Normalization of the specific protein band was performed according to creatinine content. A total of 127 samples were analysed supporting a significant difference between patients with PCa and patients with negative histology report (P < 0.001). A protein band of approximately 40 kDa corresponding to ZAG was solely detected.

A total urine volume of 15 µL was separated by 10% SDS-PAGE under reducing conditions and electroblotted to Hybond-ECL nitrocellulose membrane (Amersham Biosciences, Little Chalfont, UK). After blocking with 5% non-fat dried milk in TBST (20 mM Tris, pH 7.6, 137 mM NaCl; Applichem, Darmstadt, Germany, 0.1% Tween 20; Applichem, Darmstadt, Germany) for 2 h at room temperature, membranes were washed with TBST and incubated overnight at 4 °C with the primary antibody, mouse anti-human ZAG (Antibodies: Santa Cruz, Santa Cruz, California, USA; dilution 1:1000). Membranes were then washed with TBST and incubated with anti-mouse horseradish-peroxidase-conjugated secondary antibody (Santa Cruz Biotechnology; dilution 1:2000) for 2 h at room temperature. A final wash with TBST was made and target protein was detected with the enhanced chemiluminescence (ECL detection system: Perkin Elmer, Waltham, Massachusetts, USA) detection system. Films were scanned and images were analysed using Quantity One software (Bio rad, Hercules, California, USA). Normalization was made according to the total amount of creatinine that was included in the 15 µL urine samples.


Immunohistochemical staining was performed based on the EnVision+ System-HRP (DakoCytomation). Formalin-fixed paraffin-embedded blocks from nine randomly selected cases (six PCa Gleason's 3+3, score 6, and three PCa Gleason's 4+4, score 8) were cut into 4 µm thick sections and placed on polylysine-coated glass slides. Tissue sections were deparaffinized, rehydrated and incubated in 0.01 M citrate buffer pH 6 for 30 min in a microwave oven at 800 W, treated with 3% hydrogen peroxide for 15 min and rinsed. After cooling for 20 min, they were incubated overnight at 4 °C with the primary anti-ZAG polyclonal antibody (F22, sc-11242, Santa Cruz Biotechnology, dilution 1 : 5000) and then incubated for 30 min with the anti-mouse horseradish-peroxidase-labelled polymer (EnVision+ System-HRP). Finally, sections were treated with a diaminobenzidine chromogenic substrate (BioGenex) for 10 min, counterstained with haematoxylin, dehydrated and coverslipped.

Statistical analysis

Continuous variables are presented as mean and standard deviation or median and interquartile range (IQR), while qualitative variables are presented as absolute and relative frequencies. Receiver operating characteristic (ROC) curves were used to estimate the predictive ability of ZAG. Also, ROC curve analysis determined the optimal sensitivity and specificity by using various cut-off values for the prediction of PCa. Afterwards, in order to estimate ZAG's additional predictive ability, logistic regression models were used to derive linear predictors and compare the areas under the curve (AUCs). A stepwise multiple logistic regression analysis (P for removal was set at 0.1 and P for entry was set at 0.05) was performed to identify independent predictors for PCa. Adjusted odds ratios with 95% confidence intervals were computed from the results of the logistic regression analyses. Model diagnostics were evaluated using the Hosmer and Lemeshow statistic. All P values reported are two-tailed. Statistical significance was set at 0.05 and analyses were conducted using STATA statistical software (version 9.0).


Sample baseline characteristics are presented in Table 1.

Table 1. Demographics and clinical characteristics
  N (%)
  1. BMI, body mass index; PIN, prostatic intraepithelial neoplasia.

Age (years), mean (sd)65.7 (8.7)
BMI (kg/m2), mean (sd)27.5 (3.5)
 No44 (34.6)
 Ex-smoker48 (37.8)
 Yes35 (27.6)
 Pack years, median (IQR)0 (0–24)
Family history of cancer 
 No60 (47.2)
 Yes67 (52.8)
Family history of prostate cancer
 No102 (80.3)
 Yes25 (19.7)
PSA (ng/mL), mean (sd)9.1 (5.3)
PSA (ng/mL) 
 <46 (4.7)
 4–1089 (70.1)
 >1032 (25.2)
 Negative68 (54.0)
 Positive58 (46.0)
 Without cancer56 (44.1)
 PIN29 (22.8)
 Prostate cancer42 (33.1)
ZAG, median (IQR)1.2 (0.5–2.1)

ROC curve analysis (Table 2) showed a significant predictive ability of ZAG for PCa. The AUCs for the prediction of PCa were 0.68 (95% CI 0.59–0.78) for ZAG. Also, PSA had a significant predictive ability for PCa with AUC equal to 0.65 (95% CI 0.54–0.75, P= 0.007). ROC curve analysis (Fig. 2a) showed that the optimal-cut off of ZAG for the prediction of PCa was 1.13 with sensitivity equal to 78.6% and specificity equal to 60%. When considering ZAG and PSA together there was a significant improvement in the predictive value (P= 0.010) with AUC equal to 0.75 (95% CI 0.66–0.85) (Fig. 2b).

Table 2. ROC analysis for ZAG
 AUC95% CI P
PSA 4–10 ng/mL0.680.56–0.790.012
Negative DRE0.690.56–0.820.028
Figure 2.

a, ROC curve for the prediction of PCa from ZAG in the total sample. b, ROC curve for the prediction of PCa from PSA alone and PSA with ZAG in the total sample. c, ROC curve for the prediction of PCa from PSA alone and PSA with ZAG in patients with PSA 4–10 ng/mL.

Separate analysis of patients with PSA levels 4–10 ng/mL showed that ZAG had a discriminative power with AUCs equal to 0.68 (Table 2). The optimal cut-off was 1.13 for ZAG with sensitivity equal to 73.9% and specificity equal to 59.1%. In patients with PSA levels 4–10 ng/mL, ZAG improved the predictive ability of the model significantly (P= 0.05) (Fig. 2c). In the group of patients with negative DRE, ZAG had discriminative power with AUC equal to 0.69, but did not improve the predictive power of the model with PSA (P= 0.167).

When multiple logistic analysis in a stepwise method was conducted with dependent variable the presence of PCa it was found that ZAG values more than 1.13 were independently associated with PCa. Greater age, greater PSA and positive DRE were associated with greater likelihood for PCa, as expected. Additionally, an optical density of ZAG of more than 1.13 had 6.88 times greater odds for PCa adjusted for the other predictors (Table 3). The Kruskal–Wallis test was used for the comparison of ZAG between different patient groups (Table 4). ZAG levels were significantly increased as PSA was increased (P= 0.005) (Fig. 3a). Also, ZAG levels were significantly more increased in the patients with positive histology for PCa (P= 0.004) (Fig. 3b). ZAG levels were not significantly associated with Gleason score (P= 0.291). No significant association was found between ZAG and pack years (r=−0.10, P= 0.272) or between ZAG and body mass index (r=−0.09, P= 0.330).

Table 3. Odds ratios (OR) and 95% CI derived from stepwise multiple logistic regression analysis with dependent variable the presence of PCa
 OR (95% CI) P
  • *

    Indicates reference category.

Age (years)1.08 (1.02–1.14)0.005
PSA (ng/mL)1.08 (1–1.17)0.039
 Positive2.70 (1.17–6.24)0.020
 ≥1.136.88 (2.5–18.89)<0.001
Table 4. Kruskal–Wallis test – association of ZAG with PSA levels, Gleason score, histology and smoking
  1. PIN, prostatic intraepithelial neoplasia.

Gleason score   
Histology results   
 High grade PIN0.80.4–1.4 
Figure 3.

a, Box plots for ZAG according to PSA levels. b, Box plots for ZAG according to histology.

In order to evaluate ZAG protein expression in tissue samples of PCa, we additionally performed immunohistochemistry on newly cut 4 µm sections from formalin-fixed paraffin-embedded tissue blocks from nine randomly selected cases (six PCa Gleason's score 6 and three PCa Gleason's score 8) corresponding to the clinical samples already analysed. As expected, all non-neoplastic prostatic epithelial tissue presented a strong cytoplasmic ZAG positivity. ZAG immunoreactivity in Gleason's score 6 PCas was identical to or slightly weaker than the immunoexpression of normal acini, while in Gleason's score 8 PCas ZAG protein expression was absent to focally weak (Fig. 4).

Figure 4.

Absent and focally weak cytoplasmic immunoreactivity for ZAG in a Gleason's 4+4 prostatic carcinoma. Normal acici are strongly ZAG positive (Magnification 200×).


ZAG showed a significant predictive ability for PCa. With an optimal cut-off value of the optical density for prediction of PCa at 1.13, ZAG showed sensitivity equal to 78.6% while maintaining specificity equal to 60%. The addition of ZAG to PSA improved significantly the latter's predictive ability for all patients independently of their actual PSA values.

The need to identify new biomarkers for the early detection of PCa has become more evident after the confounding conclusions of the two most important major trials concerning PCa screening, the ERSPC and the PLCO which were published simultaneously in 2009 [9–11]. It was stated that the current standard, PSA testing combined with DRE, is minimally invasive and easily available but does not seem to be ideal in reducing mortality [11]. Emerging evidence is pointing to the use of multiple markers in order to fully characterize the heterogeneity of prostate tumour phenotypes in combination with clinical and demographic data that would aid in predicting patients who are at risk for developing PCa and for assessing their prognoses.

Proteomics allows the combinatorial monitoring of multiple proteins that could be involved in oncogenesis and cancer progression and as a result allows potential biomarker panels as diagnostic and prognostic signatures of such events to be identified [7]. In one such study of ours, an in-depth relative quantitative profiling of proteins in clinical tissue specimens derived from patients with PCa vs BPH was achieved [7]. Based on its findings combined with a thorough literature research on potential PCa biomarkers, secreted proteins, such as ZAG, were deemed more useful relative to intracellular proteins given their greater propensity to be detected in the less invasive biological fluid compartments.

ZAG is a glycoprotein that is secreted by both normal and malignant prostate epithelia and it is present naturally in most body fluids, such as blood, sweat, seminal fluid, breast cyst fluid, cerebrospinal fluid and urine [12–14]. The function of ZAG in PCa is unknown. Nevertheless, recent data support a link between absent or weak ZAG expression in radical prostatectomy specimens and clinical recurrence after prostatectomy. It has been speculated that loss of ZAG expression may be associated with the progression to an aggressive, androgen-independent phenotype [8].

However, comparison between tissue and urine expression of ZAG seems to be a debatable issue [6–8]. In our previous proteomic study using tissue specimens, although based on a small sample, ZAG was downregulated (0.43-fold) in PCa. This result was in accordance with other studies using tissue samples [6,15,16]. In the present study, in order to provide an independent confirmation on the differential expression of ZAG in benign and malignant prostatic tissue and to examine the trend of ZAG levels in tissue in comparison with the results of other published studies we conducted a selective immunohistochemical evaluation [6]. Our results, although based on a very small sample, demonstrated a gradual decrease of the expression of ZAG, which was always very strong in the benign epithelium and nearly negative in Gleason pattern 4 carcinomas, showing a similar trend to the immunoexpression that was observed in the study of Hale et al.[6]. However, contrary to the downregulation of ZAG in tissues, our results on the concentration of ZAG in urine demonstrated a significant increase of ZAG levels in patients with positive histology for PCa. Moreover, ZAG showed a significant predictive ability for PCa which was improved significantly when the values of ZAG and PSA were considered together. Although we did not measure the levels of ZAG in serum the increase of ZAG levels in urine in patients with positive histology for PCa seems to be in accordance with reports in the literature for the ZAG levels in serum in patients with PCa [6,14,15].

The differential expression of ZAG indicated by proteomics analysis in tissue samples was also confirmed by western blot analysis in individual urine samples. Immunoblot analysis is a very sensitive method for protein detection. However, in some cases its accuracy suffers from non-specific binding of the antibody that is being used. In our study no such case was observed and a protein band of approximately 40 kDa corresponding to ZAG was solely detected.

There are two logical ‘pathways’ for detecting ZAG in urine: either via the systemic circulation or via direct spillage in urine after a vigorous prostatic massage. A possible theory explaining the conflicting result between urine and tissue samples could implicate the actual chemical nature of ZAG. First, ZAG can conjugate with α2-globulins, given their similar electrophoretic motility, making its efficient purification and detection more difficult [6]. Thus, it might create complexes, which are not filtered and not detected. Second, ZAG may bind to the prostatic epithelium and possibly conjugate with PSA itself, which may explain the possible false negative downregulation in serum specimens. Moreover, a ‘spillover’ phenomenon whereby the tumour cell produced ZAG protein infiltrates into the stroma and then into lymphatics and vessels may have taken place [6]. This phenomenon may explain the opposite ZAG expression results between the tissue and urine samples.

Different studies have reported that ZAG could fulfil the criterion as a potential new biomarker for PCa, both because it is implicated in an important pathogenetic way leading to cachexia and because it has different concentrations in tissue and blood of patients with PCa contributing to the discrimination from healthy individuals [6,12,13,15,16].

We chose urine as our biological fluid because it is easily obtained, with no invasiveness to the patient, and because other markers such as sarcosine, annexin A3, TMPRSS2:ERG fusion gene and PCA3 (already used in a clinical setting) have been proved useful as urine biomarkers [17–19].

The main concerns of our current study are the small number of patients (n= 127) and the fact that we have no information concerning the levels of ZAG in the urine of patients diagnosed with other types of cancer. Therefore the specificity of ZAG for the assessment of PCa can be questioned. Of course this is an ongoing study in which we have the intention to also measure ZAG in blood samples. It is important to mention that ZAG has been reported to be secreted from benign prostatic epithelia, raising concerns about the specificity of this protein for PCa, but the same issues have been addressed for PSA [6]. A concern for the specificity of ZAG raised also by Hale et al.[6] is that ZAG production by tumour-activated normal secretory epithelial cells within the prostate or elsewhere in the body may contribute to serum ZAG levels in patients with PCa. Nevertheless in our case ZAG with an optimal cut-off for the prediction of PCa at 1.13 showed specificity equal to 60%. ZAG also showed an increasing concentration in the urine of patients with higher PSA (P= 0.005) and a positive histology for PCa (P= 0.004) indicating a good predictive ability either combined with PSA or as a solo biomarker. Finally it is important to mention that although ZAG's tissue expression was absent to focally weak with higher Gleason score confirming the results of previous studies in the literature this raises the concern of missing patients with PCa with high Gleason score using ZAG levels in urine [8].

Our strengths are that this is a well designed prospective study, with well characterized biological specimens. Additionally, the final selection of ZAG was made after a combination of the interpretation of the results of our previous proteomic analysis and an analytical search of the literature concerning PCa biomarkers. Studies with a much larger number of patients are needed together with the evaluation of ZAG expression in cancers other than prostate in order to decide whether ZAG could be used in everyday clinical practice as a novel biomarker for the early diagnosis of PCa most probably as part of a panel of biomarkers.

In conclusion, ZAG detected in the urine of patients eligible for TRUS-b, thus belonging to a high risk group for PCa diagnosis, was shown to be increased and significantly associated with PCa. ZAG also increased the predictive ability of PSA when combined together. Still, its exact role in the pathogenesis of the disease remains unclear. Further validation is required in order for ZAG to be established as a novel biomarker or included in a panel of biomarkers for the early diagnosis of PCa.


None declared.