Increased Expression of Anterior Gradient-2 Is Significantly Associated with Poor Survival of Prostate Cancer Patients

Increased expression of anterior gradient-2 is significantly associated with poor survival of prostate cancer patients

Y Zhang1, SS Forootan1, D Liu2, R Barraclough3, CS Foster1, PS Rudland3 and Y Ke1 1Molecular Pathology Laboratory, School of Cancer Studies, The University of Liverpool, Liverpool, UK; 2Cancer Tissue Bank Research Centre, The University of Liverpool, Liverpool, UK and 3School of Biological Sciences, The University of Liverpool, Liverpool, UK

Anterior gradient-2 (AGR2) expression was examined in a series of prostate cell lines and in an archival set of prostate tissues. The relative levels of AGR2 expression in the malignant cell lines PC-3 and PC-3M were, respectively, 5.370.1 and 3.870.2 times that detected in the benign cell line PNT-2. Immunohistochemical staining in 106 cases showed that amongst seven normal cases, one (14.3%) was unstained, five (71.4%) stained weakly positive and one (14.3%) stained moderately positive. Amongst 34 benign prostate hyperplastic (BPH) cases, 12 (35.3%) were unstained, 18 (52.9%) stained weakly positive and four (11.8%) stained moderately positive. Amongst 65 carcinomas, three (4.6%) were unstained, 14 (21.5%) stained weakly positive, 19 (29.2%) stained moderately positive and 29 (44.9%) stained strongly positive. AGR2 expression in carcinomas was significantly higher than that in BPH (v2-test, Po0.001). Kaplan–Meier survival analysis showed that increased AGR2 expression was significantly (log rank test, P ¼ 0.007) associated with reduced patient-survival time. Increased joint Gleason score (GS) was significantly (log rank test, P ¼ 0.001) associated with poor patient survival. However, neither prostate specific antigen (PSA) level, nor androgen receptor (AR) index, was significantly associated with patient-survival time. Increased AGR2 expression was significantly correlated with high GS (two-sided Fisher’s exact test, Po0.001) and PSA levels (Mann–Whitney U-test, P ¼ 0.047), but not significantly related to the level of AR (Mann–Whitney U-test, P ¼ 0.286). These results suggest that increased AGR2 expression is a valuable prognostic factor to predict the clinical outcome of the prostate cancer patients. Prostate Cancer and Prostatic Diseases (2007) 10, 293–300; doi:10.1038/sj.pcan.4500960; published online 24 April 2007

Keywords: AGR2; androgen receptor; Gleason scores; patient survival; PSA

Prostate cancer is the most common male cancer and a characterize new molecular markers for prediction of significant health problem in industrialized countries.1–3 the patient outcomes. It was estimated that 232 090 men had been diagnosed Androgen-inducible gene anterior gradient-2 (AGR2), with prostate cancer and 30 350 died from the disease in also known as HAG-29 and Gob-4,10 is the human the year 2005.4 During the past several years, great effort orthologue of the secreted Xenopus laevis AGR protein has been made to find out those genes promoting or (XAG-2) which has been known to play a patterning role suppressing malignant progression of prostate cancer in frog embryos.11 AGR2 was originally cloned as a gene cells3,5 and many new genes involved in carcinogenesis that was differentially expressed between estrogen and metastasis have been identified.6,7 In contrast to receptor (ER)-positive carcinoma cell lines MCF-7 and rapid development in identification of new oncogenes ER-negative benign breast cell lines MDA-MB-2319 and and tumour suppressor genes, progress in searching for Hama 123.12 AGR2 is expressed at a higher level in the new prognostic and diagnostic markers from cancer- human breast cancer cell line MCF-7 than in the benign related genes has been relatively slow. In view of the cell line Hama 123. Previous investigation suggested that controversies existing in the accuracy and reliability of expression of AGR2 was significantly associated with currently used biomarkers, such as prostate specific poor outcome in patients with ER-positive breast antigen (PSA),8 it is of great importance for prostate cancer.13 In comparison to research in breast cancer, the cancer treatment and management to identify and role of AGR2 in prostate cancer was not very clear. AGR2 was highly induced by androgen in an androgen receptor (AR)-dependent manner and the expression of Correspondence: Dr SS Forootan, Molecular Pathology Laboratory, AGR2 was markedly increased at both mRNA and School of Cancer Studies, The University of Liverpool, 5/6 Floor, protein levels in prostate cancer cells compared with Duncan Building, Daulby Street, Liverpool L69 3GA, UK. 14 E-mail: [email protected] their benign counterparts, indicating that AGR2 ex- Received 20 November 2006; accepted 12 January 2007; published pression might be involved in the malignant progression online 24 April 2007 of prostate cancer. AGR2 expression in prostate cancer Y Zhang et al

294 In this study, we aimed to find out whether AGR2 can b-actin bands. Results were expressed as the mean be used as a prognostic marker for prostate cancer (7s.d.) of four separate measurements. diagnosis. The expression status of AGR2 in different prostate cell lines and in an archival set of 106 prostate tissues has been examined. The relationship between Human prostate tissues increased expression of AGR2 and the time of patient The human prostatic tissues were an archival set with survival has been carefully assessed, and the expression up-to-date records held within the surgical pathology of AGR2 has also been correlated with Gleason score (GS) archive in the Molecular Pathology Laboratory. Tissues of the carcinomas, levels of PSA and AR. were taken from 65 prostate adenocarcinoma patients with a mean age of 73 years and from 34 benign prostate hyperplastic (BPH) patients with a mean age of 67.5 years through Trans-urethral Resection of Prostate in the Materials and methods Royal Liverpool University Hospital during the 5-year period of 1995–2000. The seven normal prostate tissues Cell lines and culture conditions were taken from road traffic accident victims (mean age Five human prostate cell lines were used in this work: 48 years) who did not have a history of any prostatic the benign prostate epithelial cell line PNT-2,15,16 the diseases. Our study was performed in accordance with weakly malignant cell lines LNCaP,17 highly malignant the Medical Research Council guidelines and was cell lines Du-145,18 PC-3 and PC-3M.19 The cells were approved by Liverpool Local Science Ethics Committee cultured as monolayers in RPMI 1640 (Invitrogen, (project reference number: Ke: 02/19). Four-micrometer sections of formalin-fixed and paraffin-embedded tissues Paisley, UK) supplemented with 10% (v/v) FCS (Biosera, 21,22 East Sussex, UK), penicillin (100 units/ml), streptomycin were cut and processed, as described previously. (100 mg/ml) (Invitrogen), hydrocortisone (50 ng/ml) and Tissue sections were examined independently by two testosterone (50 ng/ml) (Sigma, Grillingham, UK). qualified observers which were blinded to the clinical data and classified as normal, BPH and carcinomas. Carcinomas were further classified according to their GS.23 PSA level at the initial diagnosis was obtained Detection of AGR2 protein in cell lines through telepath system and classified into two groups AGR2 protein expressed in various cell lines was of low (o4–10 ng/ml) and high (410 ng/ml).24 detected by western blot, using an ECL light-emitting non-radioactive kit (Amersham Bioscience, Bucks, UK) as described previously.20 Proteins were extracted from Detection of protein expression in prostate tissues different cell lines and equal amounts of each sample AGR2 expression in human prostate tissues was detected (20 mg) were quantified with a Coomassie Protein Assay by immunohistochemical staining. The procedures for Reagent kit (Biorad, Hemel, Hempstead, UK). Samples tissue staining were similar to those described pre- containing equal amounts of total protein were subjected viously.6,22 After dewaxing in xylene and rehydration in to SDS–PAGE in 12% (w/v) polyacrylamide gels together graded ethanol, sections were immersed in methanol with molecular weight markers. Recombinant AGR2 containing 3% (v/v) hydrogen peroxide for 12 min to protein flanked with a histidine tag containing a protease block endogenous proxidase, followed by rinsing in factor X cleavage site was produced in Escherichia coli, running tap water and immersing in distilled water. purified as described previously12 and loaded on the left Antigen retrieval was then achieved with pressure- side of the gel to act as a size control. The separated cooker pretreatment for 3 min. Sections were incubated proteins were transferred onto a nitrocellulose mem- within house (12) affinity-purified AGR2 antibody with a brane (Hybond, Amersham Biosciences) at 100 V for 2 h 1/500 dilution for 1 h. Bound antibodies were detected at 41C. The membrane was first incubated with pre- with 100 ml of reagents from Envision System Horse- blocking reagents (ProtoBlock, National Diagnostics, radish Peroxidase kit (Dakocytomation, Ely, UK) for Atlanta, USA) for 1 h at room temperature and then 30 min and the reaction was visualized with DAB (3–30- incubated with affinity-purified, in-house rabbit poly- diaminobenzidine) for 10 min. Afterward, sections were clonal anti-human AGR2 antibody12 with 1/500 dilution briefly counterstained with hematoxylin and mounted overnight at 41C. The membrane was then incubated with dibutyl phthalate xylene (DPX). To test its specifi- with polyclonal swine anti-rabbit immunoglobulins/ city, the antibody was incubated with 10 mg/ml of HRP (DAKO, CAMBS, UK) with 1/1000 dilution for recombinant human AGR2 protein at 41C overnight 1 h at room temperature and bound antibodies were and applied to sections as described above; this revealed by chemiluminescene (ECL, Amersham Bio- procedure completely blocked any immunolabelling. Tech) and recorded on Kodak XAR-5 film. The bands ARs were detected using monoclonal anti-human AR were then scanned using Alpha-Imager 2000 software antibody (Dako) on the same set of tissues at a 1/100 (Alpha Innotech, Cannock, UK) and the intensity of band dilution of the primary antibody following the pressure- was obtained by measuring peak area. The relative cooker pretreatment. levels of AGR2 expression in different cell lines were Immunoreactive scores of AGR2 were established by obtained by comparing the intensity of bands. The level the method described previously25 with some modifica- of AGR2 expressed in benign cell line PNT2 was set tions. Stains were first assessed by the percentages of at 1. To standardize the quantitative measurements, the cells stained to obtain a percentage score (1–4) [1 (o5%), blot was incubated with specific anti-human actin 2 (5–30%), 3 (31–70%), 4 (470%)); then by intensity of the monoclonal antibody (Sigma) and possible loading stains to obtain an intensity score (1–4) (1 (À), 2 ( þ ), 3 errors were corrected by relating the images to the ( þþ), 4 ( þþþ)]. The final scores (1–16) were obtained

Prostate Cancer and Prostatic Diseases AGR2 expression in prostate cancer Y Zhang et al 295 by multiplying the percentage score and the intensity a 1 2 3 4 5 kDa score. The immunohistochemical stains were finally classified as negative (1), weakly positive (2–6), moderately positive (7–11) and strongly positive (12–16). To assess the staining index of AR, each slide was scored by AGR2 20 the intensity of staining and percentage of positive cells (labelling frequency, %). The grading scale for intensity ranged from no signal (0) to strong signal (3); grade 2 corresponded to a moderate signal seen in low-to- β-actin 42 intermediate power and grade 1 corresponded to weak signal in intermediate and high power. The labelling 6 frequency was defined as 0 (0%), 1(1–30%), 2 (31–70%) b and 3 (70–100%). AR index was obtained by multiplying 5 the grade of intensity by the labelling frequency for each 4 case. AR expression has been defined as low (o4) and 3 X 26 high ( 4). 2 1

Statistical analysis of AGR2 level Relative 0 Statistical analyses were performed using the SPSS 12345 package, version 13.0 (SPSS Inc., Chicago, IL, USA). Figure 1 Measurement of the levels of AGR2 expression in benign Association between the AGR2 expression and the nature and malignant prostate cell lines. (a) Western blot analysis of AGR2 of the prostate tissues (normal, benign and malignant) expression in different prostate epithelial cell lines. Lane 1, the was assessed by two-sided Fisher’s exact test. Correla- benign cell line PNT-2; Lane 2, the weakly malignant cell line tions between AGR2 expression, PSA level, AR expres- LNCaP; Lanes 3, 4 and 5, the highly malignant cell lines Du-145, PC- 3 and PC-3M. The antibody against b-actin was used to standardize sion and GS in prostate carcinoma and patient-survival the amount of proteins loaded in each sample on the same blot. (b) time were evaluated by using Kaplan–Meier plots and Relative levels of AGR2 protein expressed in different cell lines. The differences between groups of patients were assessed by expression of AGR2 measured in benign PNT-2 cells was set at 1.0 the log rank test. Box plot and Mann–Whitney U-test and the relative levels of AGR2 expressed in other cell lines were were used to assess correlations between AGR2 and obtained by comparing them with the level that in PNT-2. The PSA level, AGR2 and AR and PSA levels and GS. possible loading errors were corrected by relating levels of AGR2 to the amount of b-actin. Results were the mean (7s.d.) of four P-valueso0.05 were considered statistically significant. separate measurements.

unstained, 18 (52.9%) stained weakly positive and four Results (11.8%) stained moderately positive (Table 1). The difference of AGR2 stains between normal and BPH AGR2 expression in prostate cell lines cases was not statistically significant (two-sided Fisher’s Western blot analysis detected a single AGR2 band with a exact test, P41). Amongst 65 carcinoma cases, three size of 20 kDa in the benign PNT-2 cell line and in highly (4.6%) were unstained, 14 (21.5%) stained weakly malignant cell lines, PC-3 and PC-3M (Figure 1a). The positive, 19 (29.2%) stained moderately positive, and 29 detailed quantitative measurements of relative AGR2 (44.9%) stained strongly positive (Table 1). The staining levels expressed in different cell lines are shown in in carcinoma cases for AGR2 was significantly higher Figure 1b. When the level of AGR2 expressed in the than those observed in normal and BPH cases (two-sided PNT2 cells was set at 1, the relative levels of AGR2 in the Fisher’s exact test, Po0.001). malignant cell lines PC3 and PC3M were 5.2970.14 and 3.8170.23 (mean7s.d.), respectively. No AGR2 protein was detected in the weakly cell lines LNCaP and highly AGR2 expression and patient survival malignant Du145. The relationship between the strength of AGR2 staining and the patient-survival time are shown in Figure 3a. For the patients with strongly positive AGR2 stains, the AGR2 expression in prostate tissues median survival time was 36.3 months, which was Examples of immunohistochemical stains of different significantly shorter than 75 months and 80 months for prostate tissues with rabbit polyclonal anti-human AGR2 patients with moderately (log rank test, P ¼ 0.03) and antibody were shown in Figure 2. In normal prostate weakly (log rank test, P ¼ 0.006) positive AGR2 stains, gland, AGR2 staining was localized predominantly to the respectively. Overall survival time was significantly (log cytoplasm of luminal cells of prostate glands, whereas in rank test, P ¼ 0.007) reduced with increased AGR2 stains. the prostate carcinomas, both cytoplasmic and nuclear The difference in median survival time between moder- stains were observed. When the antibody was neutra- ately and weakly positive AGR2 staining groups was not lized with the recombinant AGR2 protein, the stain was significant (log rank test, P ¼ 0.25). completely blocked (Figures 2g and h), indicating the specificity of the antibody for AGR2. Amongst seven normal prostate tissues, one (14.3%) was unstained, five GS and patient survival (71.4%) stained weakly and one (14.3%) stained moder- To assess the relationship between the GS and ately positive. Amongst 34 BPH cases, 12 (35.3%) were patient survival, 65 carcinoma cases were divided

Prostate Cancer and Prostatic Diseases AGR2 expression in prostate cancer Y Zhang et al 296

Figure 2 Detection of AGR2 expression in normal, benign and malignant prostate tissues by immunohistochemical staining. (a) Normal prostate exhibiting either negative or weakly positive stain ( Â 100). (b) BPH sample stained weakly positive ( Â 100). (c) Weakly malignant carcinoma (GS 4) stained weakly to moderately positive ( Â 100). (d) Moderately malignant carcinoma (GS 6) stained strongly positive ( Â 100). (e) A highly malignant carcinoma (GS 9) stained strongly positive ( Â 100). (f) Strongly positive stain was observed in a malignant carcinoma (large arrow) whereas no stain was observed in the adjacent benign areas (small arrow) within the same tissue sample ( Â 100). (g) and (h) are test stains for antibody specificity: the moderately malignant carcinoma stained strongly positive (g); but when the recombinant AGR2 protein was added to neutralize the polyclonal antibody in the immunohistochemical reaction, no stain (h) was observed on the same carcinoma sample ( Â 100).

into the weakly malignant group with GS of p5, P ¼ 0.002) associated with the increased degree of the moderately malignant group with GS of 6–7 and malignancy. the highly malignant group with GS of 8–10. The median survival time of patients with weakly, moderately and highly malignant carcinomas was 80, 48 PSA and patient survival and 24 months, respectively (Figure 3b). The reduced The correlation between PSA level and patient-survival survival time was significantly (log rank test, time is shown in Figure 3c. The median survival time for

Prostate Cancer and Prostatic Diseases AGR2 expression in prostate cancer Y Zhang et al 297 Table 1 AGR2 expression in different prostate tissues patients with low (o4 ng/ml) and borderline (4–10 ng/ ml) PSA level was 60.9 months compared with 48.3 Tissues AGR2 expression No. of cases months in patients with high (410 ng/ml) PSA levels. Negative Weak Moderate Strong Although survival time has been reduced in patients with high PSA level, the association of the patient PSA Normal 1 5 1 0 7 level with the length of the survival time was not BPH 12 18 4 0 34 significant (log rank test, P ¼ 0.108). Carcinoma (total) 3 14 19 29 65 Scoresa p5 0 13 2 0 15 Scoresa 6–7 2 0 15 9 26 AR and patient survival a Scores 8–10 1 1 2 20 24 The correlations between AR and survival are shown in Abbreviations: AGR2, anterior gradient-2; BPH, benign prostate hyperplastic. Figure 3d. The median survival time for patients with aCombined Gleason scores. low (index o4) and high (index X4) AR levels was 60.6

a b 1.0 AGR2 1.0 GS 1.00 1 2.00 2 3.00 3 0.8 0.8

0.6 0.6

0.4 0.4 Cummulative Survival

Cummulative Survival 0.2 0.2 Log Rank test, P = 0.007 Log Rank test, P = 0.002 0.0 0.0

0 20 40 60 80 100 120 0 20 40 60 80 100 120 Survival time (months) Survival time (months)

c d 1.0 PSA 1.0 AR <4-10 <4 >10 >4

0.8 0.8

0.6 0.6

0.4 0.4 Cummulative Survival Cummulative Survival

0.2 0.2

Log Rank test, P = 0.108 Log Rank test, P = 0.485 0.0 0.0

0 20406080100120 0 20406080100120 Survival time (months) Survival time (months)

Figure 3 Kaplan–Meier survival curves of prostate cancer patients. The cumulative survival was related to different levels of 4 separate parameters. (a) Different levels of AGR2 expression: Group 1, negative and weakly positive AGR2 stains (n ¼ 12); Group 2, moderately positive stains (n ¼ 24); and Group 3, strongly positive AGR2 stains (n ¼ 29), as defined in Materials and methods. (b) Different GS: Group 1, GS p5(n ¼ 15); Group 2, GS 6–7 (n ¼ 26); and GS 8–10 (n ¼ 24). (c) Different levels of PSA: Group 1, PSAo4–10 ng/ml (n ¼ 38) and Group 2, PSA410 ng/ml (n ¼ 25). (d) Different AR indices: Group 1, AR indexo4(n ¼ 24) and Group 2, AR index44(n ¼ 41).

Prostate Cancer and Prostatic Diseases AGR2 expression in prostate cancer Y Zhang et al 298 and 48 months, respectively. The time of patient survival a and the level of AR were not significantly (log rank test, 200.00 Mann-Whitney U test, Weak v Strong AGR2, P = 0.015 P ¼ 0.485) associated. Moderate v Strong AGR2, P = 0.007 * Moderate v Weak AGR2, P = 0.095 *

Correlation of AGR2 expression with GS 150.00 Amongst 15 GSp5 carcinomas, 13 (86.6%) stained * weakly positive and two (13.3%) stained moderately positive. Amongst 26 GS 6–7 carcinomas, two (7.7%) were unstained, 15 (57.6%) stained moderately positive and nine (34.6%) stained strongly positive. Amongst 100.00 24 GS 8–10 carcinomas, one (4.1%) was unstained, one PSA (4.1%) stained weakly positive, two (8.3%) stained moderately positive and 20 (83.3%) stained strongly positive (Table 1). The strengthened AGR2 staining was 50.00 significantly (two-sided Fisher’s exact test, Po0.001) associated with the increased GS of the carcinomas. *

AGR2 expression and PSA level 0.00 The correlation between levels of PSA and AGR2 1.00 2.00 3.00 expression is shown in Figure 4a. For patients with AGR2 weak, moderate and strong AGR2 stains, the median PSA level was 6.85, 6.20 and 10.3 ng/ml, respectively. Box plot b Mann-Whitney U test, 10 Weak v Strong AGR2, P = 0.286 analysis showed that the level of PSA in the patients with Moderate v Strong AGR2, P = 0.600 strong AGR2 staining was significantly higher than that of patients with moderate (Mann–Whitney U-test, 8 P ¼ 0.015) and weak (Mann–Whitney U-test, P ¼ 0.007) staining. However, the difference in patient PSA levels 6 between moderate and weak AGR2 stained groups was not significant (Mann–Whitney U-test, P ¼ 0.095).

AR index 4

AGR2 expression and AR index 2 The correlation of AR index with AGR2 expression was shown in Figure 4b. For patients with weak AGR2 stains, 0 the median AR index was 1. For patients with moderate 1.00 2.00 3.00 and strong AGR2 stains, the median AR indexes were 4. AGR2 Box plot analysis showed that the difference in patient AR indexes either between strong and moderate AGR2- c 200.00 Mann-Whitney U test, * stained groups (Mann–Whitney U-test, P ¼ 0.600), or GS 6-7 v GS 8-10, P = 0.02 * GS -5 v GS 8-10, P = 0.001 between strong and weakly stained groups (Mann– GS 6-7 v GS -5, P = 0.308 * Whitney U-test, P ¼ 0.286), was not significant. 150.00 * * * PSA and GS The correlation of PSA levels and GS was shown in 100.00 Figure 4c. For patients with GS p5, GS 6–7 and GS 8–10 carcinomas, the median PSA levels were 6.2, 7.05 and PSA (ng/ml) 27.40 ng/ml, respectively. Box plot analysis showed that 50.00 the PSA level was significantly higher in patients with GS 8–10 carcinomas than those with GS p5 (Mann– Whitney U-test, P ¼ 0.001) and GS 6–7 carcinomas 0.00 (Mann–Whitney U-test, P ¼ 0.02). However, the dif- -5 6-7 8-10 ference in patient PSA levels between the GS 6–7 and Combined Gleason Score GS p5 carcinoma groups was not significant (Mann– Whitney U-test, P ¼ 0.308). Figure 4 Box plot analysis of the correlation between and differently selected variables expressed in prostate cancer. (a) Correlation between AGR2 and PSA levels in three groups of prostate cancer patients: weak (n ¼ 12), moderate (n ¼ 24), high (n ¼ 29) levels of AGR2 and PSA. (b) Correlation between AGR2 and Discussion AR index in three groups of prostate cancer patients: weak (n ¼ 12), moderate (n ¼ 24), high (n ¼ 29) levels of AGR2 and AR index. (c) AGR2 was originally found in human breast cancer. Correlation between PSA level and GS in three groups of prostate Using suppression-subtractive hybridization, it was cancer patients: GS p5(n ¼ 15), GS 6–7 (n ¼ 26) and GS 8–10 found that 29 genes expressed in ER-positive breast (n ¼ 24). ‘J’ is the representative of outlier values and ‘*’ is the carcinomas might contribute to its less aggressive representative of extreme values.

Prostate Cancer and Prostatic Diseases AGR2 expression in prostate cancer Y Zhang et al phenotype when compared with ER-negative tumours. The overall increment of AGR2 expression was 299 The ER expression in eight breast carcinoma cell lines significantly (log rank test, P ¼ 0.007) associated with was correlated with the expression of one of the 29 genes, the reduced time of patient survival. The survival time of DEME2.27 After screening an ER-positive breast cancer the patients with strong AGR2 staining was significantly cell cDNA library with the DEME2 probe and searching shorter than that of the patients with moderate and weak expression sequence tags (EST) database, AGR2, human AGR2 staining (Figure 3a). This result showed that the homologue of the XAG-2 protein was discovered.9 It was overall increased AGR2 expression was significantly found that XAG-2 was a direct inducer of anterior neural associated with the poor prognosis in terms of patient fate of Xenopus ectoderm and its overexpression did not survival. Further assessment showed that the patient- result in activation of cephalic hedgehog nor in induction survival time (Figure 3b) was significantly (log rank test, of noggin expression.11 The deduced 175-amino acid P ¼ 0.002) correlated in a reciprocal manner to the degree soluble AGR2 protein, which is 91 and 47% identical to of malignancy classified by GS. These results suggested mouse and frog homologues, respectively, contains a that like GS of the carcinomas, the AGR2 staining signal peptide. Northern blot analysis detected AGR2 strength was also a very useful parameter in predicting transcripts of 0.9 kb in lung cancer cells and 1.6 kb in ER- the patient outcome. positive breast cancer cells; whereas a weaker expression PSA is currently the most widely used marker for was detected in pancreas. RNA dot blot analysis detected prostate cancer.34,35 In this study, although the survival a strong AGR2 expression in trachea, lung, stomach, time was reduced with increased PSA level (Figure 3c), colon, prostate and small intestine.9 By immunohisto- the increased PSA level was not significantly (log rank chemistry and real-time quantitative PCR analyses, it test, P ¼ 0.108) associated with the reduced patient- was found that AGR2 mRNA and protein exhibited survival time. Thus, in this group of patients studied, similar increase in breast cancer tissues and expression of PSA level was not a reliable prognostic marker. AR was AGR2 was correlated with expression of ER.28 In a recent believed to play an important role in regulation of the study, it was found that there was a significantly malignant progression of prostate cancer,26,36 but the progressive reduction in patient-survival time with result (Figure 3d) in this study showed there was no increasing AGR2 staining solely in ER-positive cases.13 significant (log rank test, P ¼ 0.485) association between Further analysis suggested that AGR2 was a prognostic AR index and the time of patient survival. marker in breast cancer for the subgroups of nodal When the correlations between AGR2 expression and negative stage I tumors.29 Recently, preliminary investi- PSA level (Figure 4a) and between AGR2 and AR index gations were performed on the possible role of AGR2 in (Figure 4b) were assessed, the level of PSA in the patients other diseases, such as inflammatory bowel disease30 with strong AGR2 staining was significantly higher than and prostate cancer.14,31 Although it was reported that that of the patients with moderate staining (Mann– the overexpression of AGR2 was detected in prostate cell Whitney U-test, P ¼ 0.007) and weak staining (Mann– lines and carcinomas, its actual role in malignant Whitney U-test, P ¼ 0.015), but the difference in patient progression was not clear. Its prognostic significance in PSA levels between moderate and weak AGR2 stains was patient survival has not been carefully investigated. not significant (Mann–Whitney U-test, P ¼ 0.095), sug- In this study, we first examined the AGR2 expression gesting that the level of AGR2 expression was partially status in five commonly used prostate cell lines associated with the level of PSA. In contrast, the (Figure 1). We found that two of the four (50%) malignant difference in patient AR indices was neither significant cell lines expressed 3.8 to 5.3 times more AGR2 than the between strong and moderate AGR2 stained cases benign PNT-2 cells, whereas another two malignant cell (Mann–Whitney U-test, P ¼ 0.600), nor between strong lines did not express AGR2. This result showed that the and weakly stained cases (Mann–Whitney U-test, elevated expression of AGR2 is associated with increased P ¼ 0.286). These results suggest that the expression of malignant characteristics in 50% of the cell lines AGR2 was not related to that of AR in the cases of examined, indicating that it is possible that the increased prostate cancer studied in this work, although AGR2 was AGR2 expression plays a promoting role in the develop- closely related to ERs in breast cancer.12 More study is ment and metastasis of prostate cancer cases.32,33 required to find out why AGR2 acts in such a different The immunohistochemical analyses (Table 1) showed manner in breast and prostate cancers. no significant (two-sided Fisher’s exact test, P ¼ 0.67) The assessment (Figure 4c) showed that PSA level was difference in AGR2 staining strength between normal significantly higher in patients with highly malignant and BPH cases. However, significant (two-sided Fisher’s carcinomas than those with weakly (Mann–Whitney U- exact test, Po0.001) increase in AGR2 level was observed test, P ¼ 0.001) and moderately (Mann–Whitney U-test, in carcinomas compared with normal and BPH cases. P ¼ 0.02) malignant carcinomas, whereas the difference Furthermore, AGR2 staining was significantly stronger between the moderately and weakly malignant carcino- (two-sided Fisher’s exact test, Po0.001) in moderately ma cases was not significant (Mann–Whitney U-test, malignant tissues than that in weakly malignant tissues. P ¼ 0.308). This result showed that although PSA is not Similarly, highly malignant carcinomas exhibited sig- suitable to be a prognostic marker, it can partially reflect nificantly stronger AGR2 stains than weakly (two-sided the degree of malignancy of the carcinomas in these Fisher’s exact test, Po0.001) and moderately (two-sided groups of patients. Fisher’s exact test, Po0.01) malignant carcinomas. These In conclusion, AGR2 was expressed in higher levels in results suggest that, increased AGR2 expression was some malignant prostate cell lines and tissues compared significantly associated with increased malignant with the normal and BPH cases. The increased AGR2 changes as reflected by the increased GS. Thus increased expression was significantly associated with the in- AGR2 expression may be an indicator for the malignant creased malignancy of the carcinoma and the reduced progression of prostate cancer. survival time of patients. Therefore, the increased

Prostate Cancer and Prostatic Diseases AGR2 expression in prostate cancer Y Zhang et al 300 expression of AGR2 may be a valuable prognostic marker 18 Stone KR, Mickey DD, Wunderli H, Mickey GH, Paulson DF. to predict the outcome of patients with prostate cancer. Isolation of a human prostate carcinoma cell line (DU145). Int J Cancer 1978; 21: 274–281. 19 Kaighn ME, Lechner JF, Narayan KS, Jones LW. Prostate carcinoma: tissue culture cell lines. Natl Cancer Inst Monogr Acknowledgements 1978; 49: 17–21. We thank North West Cancer Research Fund and the 20 Ke Y, Jing C, Barraclough R, Smith P, Davies MP, Foster CS. Elevated expression of calcium-binding protein p9Ka is asso- USA Army PCRP Fund for their generous support of ciated with increasing malignant characteristics of rat prostate research grants. carcinoma cells. Int J Cancer 1997; 71: 827–837. 21 Deshmukh N, Scotson J, Dodson AR, Smith PH, Ke Y, Forster CS. Differential expression of acidic and basic fibroblast growth References factors in benign prostatic hyperplasia identified by immuno- histochemistry. Br J Urol 1997; 80: 869–874. 1 Tarassoff CP, Arlen PM, Gulley JL. Therapeutic vaccines for 22 Forootan SS, Foster CS, Aachi VR, Adamson J, Smith PH, Lin K prostate cancer. Oncologist 2006; 11: 451–462. et al. Prognostic significance of osteopontin expression in human 2 Landis SH, Murray T, Bolden S, Wingo PA. Cancer statistics. prostate cancer. Int J Cancer 2006; 118: 2255–2261. CA Cancer J Clin 1999; 41: 8–31. 23 Gleason DF, Mellinger GT. Prediction of prognosis for prostatic 3 Foster CS, Cornford P, Forsyth L, Djamgoz MB, Ke Y. The adenocarcinoma by combined histological grading and clinical cellular and molecular basis of prostate cancer. BJU Int 1999; 83: staging. JUrol1974; 111: 58–64. 171–194. 24 Ito K, Yamamoto T, Ohi M, Kurokawa K, Suzuki K, Yamanaka H. 4 Jemal A, Murray T, Ward E, Samuels A, Tiwari R, Ghafoor A Free/total PSA ratio is a powerful predictor of future prostate et al. Cancer statistics. CA Cancer J Clin 2005; 55: 10–30. cancer morbidity in men with initial PSA levels of 4.1–10 ng/ml. 5 Hughes C, Murphy A, Martin C, Sheils O, O’Leary J. Molecular Urology 2003; 61: 760–764. pathology of prostate cancer. J Clin Pathol 2005; 58: 673–684. 25 Remmele W, Stegner HE. Recommendation for uniform defini- 6 Jing C, Beesley C, Foster C, Chen H, Rudland P, West D et al. tion of an immunoreactive score (IRS) for immunohistochemical Human cutaneous fatty acid-binding protein induces metastasis estrogen receptor detection (ER-ICA) in breast cancer tissue. by up-regulating the expression of vascular endothelial Pathology 1987; 8: 138–140. growth factor gene in rat Rama 37 model cells. Cancer Res 26 Stanbrough M, Leav I, Kwan PW, Bubley GJ, Balk SP. Prostatic 2001; 61: 4357–4364. intraepithelial neoplasia in mice expression an androgen 7 Jing C, El-Ghany M, Beesley C, Foster C, Rudland P, Smith P receptor transgene in prostate epithelium. Proc Natl Acad Sci et al. Tazarotene-induced gene 1 (TIG1) expression in prostate USA 2001; 98: 10823–10828. carcinomas and its relationship to the tumorigenicity. J Natl 27 Kuang WW, Thompson DA, Hoch RV, Weigel RJ. Differential Cancer Inst 2002; 94: 482–490. screening and suppression subtractive hybridization identified 8 May F, Hartung R, Breul J. The ability of the American Joint genes differentially expressed in an estrogen receptor-positive Committee on Cancer Staging System to predict progression-free breast carcinoma cell line. Nucleic Acids Res 1998; 26: 1116–1123. survival after radical prostatectomy. BJU Int 2001; 88: 702–707. 28 Fletcher GC, Patel S, Tyson K, Adam P, Schenker M, Loader J 9 Thompson DA, Weigel RJ. hAG-2, the human homologue of the et al. hAG-2 and hAG-3, human homologues of genes involved Xenopus laevis cement gland gene XAG-2, is coexpressed with in differentiation, are associated with oestrogen receptor- estrogen receptor in breast cancer cell lines. Biochem Biophys Res positive breast tumours and interact with metastasis gene Commun 1998; 251: 111–116. C4.4a and dystroglycan. Br J cancer 2003; 88: 579–585. 10 Komiya T, Tanigawa Y, Hirohashi S. Cloning of the gene gob-4, 29 Fritzsche F, Dahl E, Pahl S, Burkhardt M, Luo J, Mayordomo E which is expressed in intestinal goblet cells in mice. Biochim et al. Prognostic relevance of AGR2 expression in breast cancer. Biophys Acta 1999; 1444: 434–438. Clin Cancer Res 2006; 12: 1728–1734. 11 Aberger F, Weidinger G, Grunz H, Richter K. Anterior 30 Zheng W, Rosenstiel P, Huse K, Sina C, Valentonyte R, Mah N specification of embryonic ectoderm: the role of the Xenopus et al. Evaluation of AGR2 and AGR3 as candidate genes for cement gland-specific gene XAG-2. Mech Dev 1998; 72: 115–130. inflammatory bowel disease. Genes Immun 2006; 7: 11–18. 12 Liu D, Rudland PS, Sibson DR, Platt-Higgins A, Barraclough R. 31 Kristiansen G, Pilarsky C, Wissmann C, Kaiser S, Bruemmendorf Human homologue of cement gland protein, a novel metastasis T, Roepcke S et al. Expression profiling of microdissected inducer associated with breast carcinomas. Cancer Res 2005; 65: matched prostate cancer samples reveals CD166/MEMD and 3796–3805. CD24 as new prognostic markers for patient survival. J Pathol 13 Innes H, Liu D, Barraclough R, Davies M, O’Neill P, Platt- 2005; 205: 359–376. Higgins A et al. Significance of the metastasis-inducing protein 32 Adamson J, Morgan E, Beesley C, Mei Y, Forster C, Fujii H et al. AGR2 for outcome in hormonally treated breast cancer patients. High-level expression of cutaneous fatty acid-binding protein in Br J Cancer 2006; 94: 1057–1065. prostatic carcinomas and its effect on tumorigenicity. Oncogene 14 Zhang JS, Gong A, Cheville JC, Smith DI, Young CY. AGR2, an 2003; 22: 2739–2749. androgen-inducible secretory protein overexpressed in prostate 33 Smith P, Rhodes N, Ke Y, Foster C. Sodium channel protein cancer. Genes Chromosomes Cancer 2005; 43: 249–259. expression enhances the invasiveness of rat and human prostate 15 Cussenot O, Berthon P, Berger R, Mowszowicz I, Faille A, cancer cells. FEBS lett 1998; 423: 19–24. Hojman F et al. Immortalization of human adult normal prostatic 34 Collette L, Burzykowski T, Schroder FH. Prostate-specific epithelial cells by liposomes containing Large T-SV40 gene. antigen (PSA) alone is not an appropriate surrogate marker JUrol1991; 143: 881–886. of long-term therapeutic benefit in prostate cancer trials. Eur J 16 Berthon P, Cussenot O, Hopwood L, Le Duc A, Matkland N. Cancer 2006; 42: 1344–1350. Functional expression of SV40 in normal human prostatic 35 Freedland SJ, Partin AW. Prostate-specific antigen: update 2006. epithelial and fibroblastic cells: differentiation pattern of non- Urology 2006; 67: 458–460. tumorigenic cell lines. Int J Oncol 1995; 6: 333–343. 36 Shah RB, Mehra R, Chinnaiyan A, Shen R, Ghosh D, Zhou M 17 Horoszewicz JS, Leong S, Kawinski E, Karr J, Rosenthal H, et al. Androgen-independent prostate cancer is a heterogeneous Chu T et al. LNCaP model of human prostatic carcinoma. Cancer group of diseases: lessons from a rapid autopsy program. Cancer Res 1983; 43: 1809–1818. Res 2004; 64: 9209–9216.

Prostate Cancer and Prostatic Diseases