Cell culture media and antibiotics were purchased from CellGro (Manassas, VA, USA). Fetal bovine serum (FBS) was obtained from Atlanta Biologicals (Lawrenceville, GA, USA) and charcoal-stripped FBS (CS-FBS) was from Invitrogen (Carlsbad, CA, USA). The synthetic androgen, R1881, was purchased from Sigma-Aldrich (St. Louis, MO, USA). The subcellular fractionation kit was from Thermo Scientific (Rockford, IL, USA). Primary antibody against AR (cat. no. 06680) was from Millipore (Billerica, MA, USA) and antibody against PRL-3 (cat. no. ab50276) was from Abcam (Cambridge, MA, USA). The secondary antibody, TX-Red conjugated goat anti-rabbit IgG (cat. no. T-2767) was from Life Technologies (Carlsbad, CA, USA). Vectashield™ mounting medium containing DAPI was purchased from Vector Laboratories (Burlingame, CA, USA). The PRL-3 expression plasmid (pMLV-PRL-3) and the empty vector (pBabe-puro) were kind gifts from Dr Y. Jiang (14). Transient transfection was performed using a Lipofectamine LTX Plus kit from Invitrogen, and carried out according to manufacturer's instructions. Vector-transfected cells were harvested 24 h post-transfection and used in proliferation, migration and invasion assays.

Oncomine is a web-based data-mining platform (http://www.oncomine.org). Oncomine's gene search function was used to locate microarray studies for which gene expression data were publicly available. PRL-3 (also known as PTP4A3) gene expression was queried between normal gland and prostate adenocarcinoma. Data from large-scale microarray studies were processed by Oncomine and results obtained were presented in box plots, along with P-values, fold change and gene rank. Query results from multiple studies showed significant increases in PRL-3 gene expression in prostate adenocarcinomas.

A high-density PCa TMA (cat. no. PR2085B) from BioMax™ (Rockville, MD, USA) containing 208 tumor cores from 114 patients was used for our immunostaining and digital image analysis (DIA). These 114 tissue cores included 8 normal samples and 106 tumor samples, which were represented by 2 transitional cell carcinomas, 12 tumor-adjacent normal tissue samples and 92 prostate adenocarcinomas. Furthermore, these 106 tumor samples consisted of 15 non-malignant cores and 91 malignant cores. In the TMA, the prostate tumor cores were also stratified by their pathological landscapes, i.e., GS, GG and T-stage.

The LNCaP, PC3, DU-145 and CWR22Rv1 cells were purchased from the American Type Culture Collection (ATCC, Rockville, MD, USA). The LNCaP-SF cell line was generated in Dr Iwasa's laboratory (15) and the LAPC4 cell line was developed in Dr van Bokhoven's laboratory (16). All PCa cell lines were maintained in RPMI-1640 medium containing antibiotics (penicillin/streptomycin) and supplemented with 10% FBS. The cells were grown at 37°C in a humidified incubator containing 5% CO₂. In specific experiments, to mimic steroid hormone deprived conditions, experiments were carried out in media supplemented with 10% CS-FBS. The cells were cultured at 37°C in a humidified incubator containing 5% CO₂.

MTT assays were performed to determine cell proliferation in both the control vector- and PRL-3 vector-transduced cells. Briefly, the cells were seeded in a 96-well plate (5×10³/well) and proliferation was measured at 24–72 h. At the indicated time points, MTT dye (20 μl) (Sigma-Aldrich) was added to each well and incubation was carried out for 3–4 h at 37°C. The formazan crystals were solubilized in DMSO and optical density (OD) was measured at 540 nm using a μQuant spectrophotometric plate reader from Bio-Tek (Seattle, WA, USA). Differences in cell proliferation between the empty vector-transduced (control) and PRL-3 transduced cells (experimental) are expressed as a percentage of the control.

Boyden chambers were used to measure the effect of PRL-3 overexpression on PCa cell invasion through Matrigel coated inserts (8-μm pore). Matrigel (cat. no. 356230) was purchased from BD Biosciences (San Jose, CA, USA). Both the control and treated cells were suspended in medium containing 1% FBS and added to the upper compartment, and medium containing 10% FBS was added to the lower chamber. The cells (2×10⁵) were placed on inserts coated with 30 μl Matrigel and invasion was measured after 20 h of incubation at 37°C. Filters were fixed with methanol for 30 min and then stained with 0.2% crystal violet in 20% methanol for a further 30 min. The inserts were washed 4 times with distilled water, dried overnight, cells counted in 5 fields, and representative images were captured. Cells were counted at ×10 magnification using an Olympus IX71 microscope with a DP71 digital camera (Olympus, Tokyo, Japan).

Wound healing assays were carried out to examine the effects of PRL-3 overexpression on the migratory phenotype of PCa cells. Briefly, the cells were grown in 6-well Petri-plates to 80–90% confluency, and a 1-ml pipette tip was used to scratch the cell monolayer. At different time points (24–72 h) images of the wound were captured using a Olympus IX71 microscope with a DP71 digital camera. Wound widths were calculated by measuring the distance between 4–5 random points within the wound edges. Changes in wound widths were calculated by dividing the average wound widths observed at 12, 24 and 48 h by the average initial wound widths at the 0–h time point.

The cells were seeded in 10-cm dishes and grown until 80–90% confluent. For androgen stimulation experiments, the cells were rinsed with 1X PBS, the medium was replaced with phenol-red free RPMI-1640 with 10% CS-FBS for 24 h, following which the cells were treated with the AR-agonist, R1881 (1 nM) for 2–48 h. Cytoplasmic and nuclear extracts were obtained using a subcellular fractionation kit (cat. no. 78840; Thermo Scientific). Cell lysates were electrophoresed using 4–20% Tris/glycine SDS-PAGE gels. Resolved proteins were then transferred onto PVDF membranes (Millipore) and non-specific binding blocked with 3% BSA in 1X TBST. The blots were incubated overnight at 4°C with 1:500 dilutions of the primary antibodies against either AR or PRL-3. Subsequently, the blots were washed and incubated with the secondary antibody for 1 h. Blots were developed using enzyme chemiluminescence (ECL) and visualized with a Fujifilm LAS-400 Luminescence Imager.

The TMA sections first underwent heat-antigen retrieval using Borg decloaker buffer (cat. no. BD100) from Biocare Medical (Concord, CA, USA). Subsequently, the sections were dehydrated with xylene and then rehydrated by sequential incubation with 100, 95 and 70% ethanol, and then washed with PBS. The samples were rinsed with 0.05% saponin in PBS. Cell permeabilization was carried out by incubating the slides in 100% methanol at −20°C for 10 min. The sections were then rinsed with 0.05% saponin washing buffer and blocked with normal goat serum containing blocking buffer (cat. no. PCN500; Life Technologies). The sections were then incubated with the primary antibody at 4°C overnight and then incubated with the secondary antibody for 1 h at room temperature and then treated with the Vectashield mounting medium.

A schematic for image acquisition and analysis is provided in Fig. 1. Briefly, immunostained images were captured using a Nikon A1 Confocal Microscope (Nikon, Melville, NY, USA) and visualized under ×60 magnification under oil immersion. An A1 Airy pinhole with a Galvano scanner and 2 frames averaging was used to capture all images in z-field planes and an auto-gain filter calibration was used to process the sub-saturated images. For each core, 5 regions of interest (ROI) were selected to represent the tumor pattern. Both DAPI (nuclei) and TX-Red (PRL-3) staining were viewed simultaneously.

Digital quantification of PRL-3 staining intensity and its subcellular localization (cytoplasm vs. nuclear) were determined by using ImageJ software with a ND2 Reader plug-in (NIH, Bethesda, MD, USA), as previously described (17). Briefly, images were analyzed in split channels for DAPI and TX-Red, and a median filter was applied (3×3 pixels radius) to control for the background. Original images were converted into binary masks, based on image thresholds. The binary image of DAPI staining was subtracted from the TX-Red binary image using the ImageJ image calculator. Subsequently, two binary overlay images, representing 'true' cytoplasmic and nuclear staining, were acquired and applied to the original PRL-3-stained images to normalize the background. Five different ROIs were assessed for their integrative density values (IDV). In each core, total PRL-3 was quantified by taking the total TX-Red IDV divided by the total DAPI IDV. In addition, to determine its subcellular levels, nuclear staining intensity IDVs were divided by the IDV of cytoplasmic staining, which provided a nuclear/cytoplasmic (Nuc:Cyto) ratio (Fig. 1).

Ki67 staining of the tumor sections was carried out to determine the proliferative index of ROIs. Briefly, the slides were incubated for 1 h with blocking buffer in normal goat serum (5% goat serum, 0.3% Triton X-100 in PBS) and MOM™ solution from Vector Laboratories. The slides were then incubated overnight at 4°C with a 1:500 dilution of Ki67 antibody (cat. no. sc-23900; Santa Cruz Biotechnology, Inc., Santa Cruz, CA, USA). Subsequently, the slides were incubated for 45 min with 1:500 dilution of the biotin-conjugated secondary antibody (cat. no. BP-9200; Vector Laboratories), washed and then incubated with horseradish peroxidase (HRP)-conjugated streptavidin (Vector Laboratories) for 25 min. The slides were then incubated with diamino benzidine (DAB; Vector Laboratories) and images captured. Images were captured using a Nikon Eclipse Ci microscope with a DS-U3 digital camera (Nikon, Tokyo, Japan).

All measurement data were exported to an Excel spreadsheet for further statistical analysis, as previously described (18,19). Groups containing <4 representations (n<4) were not considered for statistical analysis. The IDV values were averaged for each core and were grouped according to the GS, GG and primary T-stage. Groups, where continuous variables did not follow a normal distribution curve, were assessed using the non-parametric rank testing. The Mann-Whitney U test was used to compare staining differences between the malignant and non-malignant cores. The Kruskal-Wallis H test was used to compare PRL-3 staining intensities among different GS and T-stage. Analyses were performed using GraphPad Prism and SPSS statistical software packages.

To determine whether PRL-3 gene expression is associated with aggressive prostate tumors, we carried out an Oncomine™ based database query (20). The analysis of the data from 4 previous reports is shown in Fig. 2. Liu et al was the first to clearly document (P<0.0001) that PRL-3 gene expression was markedly higher in tumor cores than in normal stromal areas (Fig. 2A) (21). In a later study, in which samples were analyzed on an Affymetrix microarray, Wallace et al demonstrated (P<0.0001) a >2-fold higher PRL-3 expression in tumor samples vs. normal stromal areas (Fig. 2B) (22). By comparing a large number of prostate carcinoma and normal adjacent prostate tissue specimens, Taylor et al also showed that PRL-3 was amongst the top 1% of genes upregulated in prostate tumor cells (t=6.89, P<0.0001) (Fig. 2C) (23). In a more recent study, PRL-3 expression was measured in both castrate-resistant metastatic PCa and localized prostate carcinoma, and compared with benign prostate tissue specimens (Fig. 2D). Grasso et al clearly demonstrated that PRL-3 mRNA levels were 1.7-fold higher in tumors vs. normal cores and ranked amongst the top 2% of overexpressed genes in tumor cells (t=5.9, P<0.0001) (24). Therefore, PRL-3 gene expression is strongly associated with prostate adenocarcinoma and is amongst the top 1–3% of overexpressed genes. However, little is known about the expression of PRL-3 protein in prostate tumors.

Using a validated PRL-3 antibody, we first quantified PRL-3 staining intensity in both stromal areas and tumor nodes. Confocal immunofluorescence microscopy was used to examine the PRL-3 protein levels in PCa using a TMA containing prostate tumor cores stratified by GS, GG and tumor, node and metastasis (TNM) staging. The results obtained with the GS and GG stratification are discussed herein (Fig. 3) and those correlating with TNM classifications are presented later (Fig. 6). As compared to areas containing normal glands (control), PRL-3-specific staining was increased in aggressive tumor cores (GS6, GS7 and GS8) (Fig. 3A). In the tumor cores graded as GS6 and GS7, PRL-3 overexpression was primarily observed in the cytoplasm, and the GS8-graded cores exhibited diffuse staining with both nuclear and cytoplasmic PRL-3 detection (Fig. 3A). Furthermore, as compared to the non-malignant cores (n=15), PRL-3 expression was significantly higher in the malignant tissues (n=91) (U=212, P<0.0001) (Fig. 3B). Of note, although total PRL-3 expression was not directly associated with a lower GS, several of the higher pathological cores exhibited a clearly elevated PRL-3 expression (Fig. 3C and Table I). In all malignant tissues, the mean rank analysis demonstrated a 2.65-fold increase in total PRL-3 expression, and the highest increase was observed in GS8 cores (P<0.0001). Furthermore, although both the GS9-and GS10-graded cores exhibited a higher PRL-3 expression compared to the non-malignant cores, these aggressive tumor cores consistently exhibited lower total PRL-3 levels than the GS8 cores (P<0.05). The upregulation of total PRL-3 expression also correlated well with increasing GG tumor cores. This was evident in both GG2- and GG3-graded cores (P<0.001) and particularly in the GG4 cores (P<0.0001) (Fig. 3D and Table I). Importantly, although the GS7 cores exhibited slight increases in total PRL-3 expression (primarily cytoplasmic staining) this increase was mostly associated with the less aggressive GS7 (3+4) cores (P<0.05), but not with the more aggressive GS7 (4+3) cores, which showed evidence of nuclear PRL-3 staining.

A functional role for nuclear PRL-3 has been implicated in previous studies (25,26). Thus, in this study, to further corroborate our observations on PRL-3 overexpression in clinical samples, we carried out in vitro experiments to examine PRL-3 expression and function in different PCa cell lines (Fig. 4). Immunoblot analysis documented both the PRL-3 and androgen receptor (AR) protein levels in the androgen-dependent cell line, LNCaP, as well as in four castration resistant PCa (CRPC) cell lines, which consisted of both AR-positive CRPC lines (i.e., CWR22Rv1 and LAPC4) and AR-negative CRPC lines (i.e., PC3 and DU145) (Fig. 4A). Of note, the expression of PRL-3 directly correlated with the AR expression status of the PCa cell lines, LNCaP, 22Rv1 and LAPC4. Negligible PRL3 expression was observed in the AR-negative PC3 and DU145 cells (Fig. 4A). We then examined whether androgen (R1881) stimulation alters PRL3 levels in PCa cells. In the LNCaP cells, both AR and PRL-3 were sequestered in the cytosolic fraction (Fig. 4B, top panel) and stimulation with the AR agonist, R1881 (1 nM), caused a rapid increase in nuclear AR levels within 2 h post-stimulation. Of note, R1881 stimulation similarly increased the nuclear levels of PRL-3 in the LNCaP cells (Fig. 4B, bottom panel). The CRPC cell line, LNCaP-SF, exhibited constitutive nuclear AR levels and R1881 stimulation only slightly increased the nuclear AR levels in these cells (Fig. 4C). This clearly indicated that PRL-3 expression correlated with the AR status of PCa cell lines and that PRL-3 translocated to the nucleus following R1881 stimulation. Most interestingly, similar to AR, the aggressive CRPC cells exhibited constitutive nuclear PRL-3 levels.

We also examined whether PRL-3 overexpression alters cellular functions associated with aggressive PCa. For these experiments, the AR-null and PRL-3-null cell line, PC3, was transfected with a PRL-3 expression vector and the resultant effects on cell growth, invasion and migration were measured. Data from MTT assays revealed that, as compared to the vector controls, the PRL-3-overexpressing PC3 cells exhibited an enhanced rate of proliferation at both 48 and 72 h post-transfection (Fig. 4D). Almost a 2-fold increase in cell numbers was evident at 72 h in the PRL-3-overexpressing cells. In addition, the effects of PRL-3 overexpression on the migratory (Fig. 4E) and invasive (Fig. 4F) potential of the PC3 cells were assessed by scratch wound and Boyden chamber assays, respectively. The ectopic expression of PRL-3 did not affect the migratory behavior of the PC3 cells (Fig. 4E); however, it profoundly increased the invasive potential of these cells. As compared to the cells transfected with the control vector, the PRL-3 vector-transfected cells exhibited a 2-3-fold higher invasive potential through Matrigel-coated membrane inserts (Fig. 4F). Thus, PRL-3 overexpression may play a direct role in both the proliferation and invasion of PCa cells. These findings also suggested that the androgen-stimulated nuclear localization of PRL-3 may be associated with the aggressive phenotype of PCa cell lines.

By the digital quantification of immunostained tumor cores (Fig. 1), we evaluated the differences in cytoplasmic and nuclear PRL-3 staining (Fig. 5). The nuclear/cytoplasmic ratios for PRL-3 (Nuc:Cyto) were digitally quantified and compared with different pathological states (Table II). The GS6 tumor cores exhibited a very sparse number of cells with nuclear PRL-3, and the more aggressive tumors, particularly the GS8-and GS9-graded cores, clearly demonstrated a larger population of cells with intense nuclear PRL-3 staining (Fig. 5A). Of note, in the GS7- and GS8-graded cores, we also noted that nuclear PRL-3 was primarily localized in the central part of the nucleus. However, in the GS9-graded cores, nuclear PRL-3 was more frequently associated with the periphery of the nucleus rather than within the central regions. The analysis of both non-malignant (n=15) vs. malignant cores (n=91) clearly revealed that nuclear PRL-3 was positively associated with prostate adeno-carcinoma (U=136.5, P<0.0001). Importantly, the nuclear/cytoplasmic ratio assessments demonstrated increased (P<0.05) nuclear PRL-3 in the highly aggressive GS7 (4+3)-graded cores compared to the less aggressive GS7 (3+4)-graded cores (Fig. 5C and Table II). Similarly, despite decreases in total PRL-3 in both the GS9- and GS10-graded cores (Fig. 3C and Table I), the number of cells containing nuclear PRL-3 was significantly (P<0.0001) higher in the highly aggressive tumor cores (Fig. 5C and Table II). When grouped according to tumor grade, mean rank analysis also revealed the increased nuclear accumulation of PRL-3 in the GG4- and GG5-graded cores (P<0.0001). Nuclear PRL-3 expression exhibited a strong association with both intermediate (int) and poorly differentiated (PD) tumor patterns, as compared to well-differentiated (WD) tumors (P<0.0001) where PRL-3 was primarily localized in the cytoplasm of WD tumors (GS ≤6); however, both the intermediate (GS7) and poorly-differentiated (GS >7) tumors exhibited a higher nuclear PRL3 expression (Fig. 5D). Thus, the Nuc:Cyto ratio of PRL-3 may be used to distinguish between indolent and aggressive disease.

T-staging in the PCa specimens allows clinicians to estimate the extension of tumor burden (27,28). In this study, the TMA cohort was largely composed of T2-stage tumors and not T3-and T4-stage tumors, which made the statistical correlation more difficult. However, a trend that distinguishes the aggressive phenotype was clearly documented (Fig. 6). A significant difference in PRL-3 protein levels between non-malignant areas and the T2 cores was observed (P<0.0001). Although the T2-stage cores exhibited detectable PRL-3 expression, much stronger expression was observed in specific areas within the T3- and T4-stage tumor cores (Fig. 6A). Of note, these PRL-3-overexpressed areas were also found to contain higher levels of nuclear PRL-3 expression, and this phenomenon was particularly striking in the T4-stage tumor cores. Furthermore, as compared to the T2-stage cores, total PRL-3 expression was decreased (P<0.05) in the T4-stage cores (Fig. 6B). When comparing cores stratified by their primary tumor staging, distinct differences in both global and nuclear PRL-3 expressions were also documented. The T2-stage cores consistently exhibited a higher total PRL-3 expression as compared to the control sections (Fig. 6B and Table I). In addition, the nuclear:cytoplasmic mean ranks were significantly (P<0.0001) higher in both the T2- and T4-stage (P<0.01) cores, as compared to the control sections (Fig. 6C and Table II). Thus, nuclear PRL-3 is a reliable biomarker of clinically relevant disease and correlate with TNM classifications.

Our in vitro experiments on PRL-3-overexpressing PCa cells indicated its role in increasing cell proliferation (Fig. 4). Therefore, in prostate tumor sections, we wished to determine whether tumor cores with a high PRL-3 expression and increased nuclear PRL-3 levels are similarly associated with the proliferation marker, Ki67 (Fig. 7). Tumor cores were immunostained for both Ki67 and PRL-3, followed by DIA and comparison of the proliferative index. Pearson's r correlation analysis indicated that tumor sections with a higher nuclear PRL-3 expression were positively associated with increased Ki67-labeled cells (r=0.496, P<0.05). Similar to our findings of the total and Nuc:Cyto levels of PRL-3 in the GS-, GG- and TNM-stratified TMA cores, it was suggested that nuclear PRL-3 was directly associated with aggressively increasing areas of the prostate tumors.