Proteomic analysis of urinary extracellular vesicles from high Gleason score prostate cancer

Extracellular vesicles (EVs) are microvesicles secreted from various cell types. We aimed to discover a new biomarker for high Gleason score (GS) prostate cancer (PCa) in urinary EVs via quantitative proteomics. EVs were isolated from urine after massage from 18 men (negative biopsy [n = 6], GS 6 PCa [n = 6], or GS 8–9 PCa [n = 6]). EV proteins were labeled with iTRAQ and analyzed by LC-MS/MS. We identified 4710 proteins and quantified 3528 proteins in the urinary EVs. Eleven proteins increased in patients with PCa compared to those with negative biopsy (ratio >1.5, p-value < 0.05). Eleven proteins were chosen for further analysis and verified in 29 independent urine samples (negative [n = 11], PCa [n = 18]) using selected reaction monitoring/multiple reaction monitoring. Among these candidate markers, fatty acid binding protein 5 (FABP5) was higher in the cancer group than in the negative group (p-value = 0.009) and was significantly associated with GS (p-value for trend = 0.011). Granulin, AMBP, CHMP4A, and CHMP4C were also higher in men with high GS prostate cancer (p-value < 0.05). FABP5 in urinary EVs could be a potential biomarker of high GS PCa.

iTRAQ Analysis. We performed shotgun proteomics of EVs in urine collected after prostate massage to identify potential biomarker candidates for GS prostate cancer. In total, 18 samples (negative: n = 6; GS 6: n = 6; GS 8-9: n = 6) were labeled with iTRAQ (isobaric tag for relative and absolute quantitation) and analyzed with liquid chromatography-tandem mass spectrometry (LC-MS/MS). Patient characteristics are summarized in Table 1. A total of 4710 unique proteins were identified, from which 3528 unique proteins were quantified using 6 iTRAQ analysis sets. Gene ontology (GO) cellular component analysis showed that the most abundant proteins that could be derived from EV proteins were plasma membrane proteins (24.8%) (Fig. 2). Eleven proteins increased in patients with PCa compared to those with negative biopsy (ratio > 1.5, p-value < 0.05). The iTRAQ results of these 11 biomarker candidates are summarized in Table 2, and the iTRAQ results of all quantified proteins are summarized in Supplemental Table. Verification of Biomarker Candidates by SRM/MRM. We selected 11 proteins for verification with SRM/MRM. The selection criterion was an iTRAQ ratio of cancer/negative biopsy of > 1.5 (p-value < 0.05) ( Table 2). Twenty-nine urine samples collected after prostate massage (11 urine samples from men with negative biopsy and 18 urine samples from men with prostate cancer [5 from men with GS 6, 7 from men with GS 7, and 6 from men with GS 8-9]) were used for verification as the independent cohort. Three patients with GS 8-9 had distant metastasis of prostate cancer. Among the 11 candidate proteins, only fatty acid binding protein 5 (FABP5) was significantly overexpressed in men with prostate cancer compared with men with negative biopsy (p-value = 0.009). Moreover, FABP5 was significantly associated with GS (p-value for trend = 0.011) (Fig. 3). The receiver-operator characteristics (ROC) curve analysis showed that the area under the curve (AUC) for the prediction of prostate cancer with GS ≥ 6 by FABP5 was 0.757 (95% confidence interval [CI] 0.570-0.944, p-value = 0.027), whereas the AUC value was 0.593 (95% CI 0.372-0.815, p-value = 0.42) for prediction by serum PSA (Fig. 4A). Granulin and alpha-1-microglobulin/bikunin precursor (AMBP) were also significantly associated with GS (p-value for trend = 0.011 and 0.014, respectively), whereas neither protein was significantly overexpressed in men with prostate cancer compared with men with negative biopsy (p-value = 0.90 and 0.59, respectively) (Supplemental Fig. 1). Charged multivesicular body protein 4a (CHMP4A), charged multivesicular body protein 4c (CHMP4C), and charged multivesicular body protein 2b (CHMP2B) were overexpressed in men with GS 8-9 (p-value = 0.016, 0.042, and 0.053, respectively), whereas CHMP4A, CHMP4C, and CHMP2B were not significantly overexpressed in men with prostate cancer compared to men with negative biopsy (p-value = 0.83, 0.70, and 0.67, respectively) (Supplemental Fig. 2). Next, we determined whether FABP5 in urinary EVs after massage could predict prostate cancers with GS of 7 or more, which requires definitive treatment, in the cohort without metastasis (n = 26). Univariate logistic analysis showed that FABP5 was significantly associated with prostate cancers with GS 7 or more (p-value < 0.001). Even after adjusting for age, PSA, and PSA density, FABP5 was significantly associated with prostate cancer with GS 7 or more (p-value = 0.003) ( Table 3).

FABP5 Expression in Prostate Cancer.
We studied FABP5 expression in prostate cancer cell lines and prostate cancer tissues. Western blot analysis showed that FABP5 was expressed in EVs from the supernatants of    Table 3. Stepwise logistic regression analysis of variables associated with Gleason score ≥ 7.
PC3 and DU145 prostate cancer cell lines (Fig. 5A). Immunohistochemical analysis of prostatectomy specimens from the patients with prostate cancer showed that prostate cancer cells with Gleason pattern 4 were stained with anti-FABP5 antibody in contrast to the negative normal epithelium (Fig. 5B). These results suggested that the prostate cancer cells could be the origin of EVs with FABP5.

Discussion
The development of new biomarkers for high-risk prostate cancer is continuing by many researchers using various methods. In the prostatectomy specimens, expressions of proteins and mRNA were used to predict the prognosis of prostate cancer patients. Immunohistochemical analysis showed that the expression of annexin II in prostatectomy specimens was associated with the Gleason score 3 . The expression of pre-B cell leukemia homeobox transcription factor was associated with cancer-specific survival in patients who underwent prostatectomy 5 . Quantitative phase imaging with prostatectomy specimens could predict prostate cancer recurrence 4 . In serum, the monocyte fraction was associated with the Gleason score 6 . Plasma prostate-specific EV concentrations, which were measured by an ELISA kit for prostate-specific membrane antigen, differed in patients with benign prostatic hypertrophy and low-, intermediate-, and high-risk prostate cancer 17 . In the urine, glycosylated PSA was associated with the Gleason score 8 . " EVs were first reported in 1983 23 , and with recent advances in technology, EVs are now emerging as a resource for biomarker discovery in several types of cancer. In prostate cancer, several miRNAs (such as miR-21 and miR-107) and mRNA (AGR2 splice variant, PCA3, and TMPRSS-ERG) have been reported to be biomarkers in urinary EVs 11 .
The results of proteomic analysis of EVs in urine collected after prostate massage from men with prostate cancer were previously reported, but the number of detected proteins was up to 900 proteins 24 . Overbye et al. also reported the proteomic analysis of EV in urine collected without prostate massage 25 . They compared the urinary EVs from normal volunteers and men with prostate cancer and identified 1644 proteins in urinary EVs, but their urine samples were collected without prostate massage, which might have resulted in the detection of different urinary markers for prostate cancer. In the present study, we quantified 3528 proteins by deep proteomic analysis, and, to our knowledge, this is the largest number of EV proteins ever detected in urine from patients with prostate cancer. However, only FABP5 was significantly overexpressed in EVs in men with prostate cancer compared to men with negative biopsy in the verification cohort. This might be due to the small number of men in the verification cohort or that the limit of biomarker discovery in urinary EVs was reached with the current technique.
FABP5, also called C-FABP5 and E-FABP5, is a lipid binding protein that is found in epidermal cells 26 . FABP5s play roles in fatty acid uptake, intracellular transport, and metabolism. FABP5 may also play a role in stimulating tumor progression by transferring ligands to nuclear PPARβ /γ 27 . FABP5 promotes cell proliferation and invasion in cholangiocarcinoma 28 . Proteomic analysis of prostate cancer and benign tissue showed that FABP5 was upregulated in prostate cancer tissues 29 . Another proteomic study of surgical specimens of prostate cancer showed that tissues from lymph node metastasis had increased expression of FABP5 compared with localized prostate cancer tissues 30 . Interestingly, patients with lymph node metastasis had higher levels of serum FABP5, as measured by ELISA 30 . FABP5 promotes the proliferation of endothelial cells 31 . The present study showed the high expression of FABP5 in EVs from patients with high GS prostate cancer. FABP5 in EVs from high GS prostate cancer may affect the endothelial cells in tumor microenvironments or in distant areas and promote the metastasis of prostate cancer.
GRN, also known as granulin-epithelial precursor, is the pluripotent growth factor regulating inflammation and tumorigenesis. GRN is reported to be overexpressed in high-grade prostatic intraepithelial neoplasia and invasive prostate cancer compared with benign prostate epithelium 32 and to stimulate the migration, invasion, and proliferation of prostate cancer 33 . GRN-overexpressed hepatocellular carcinoma (HCC) cells had stem cell-like properties, and patients with GRN-overexpressed HCC had a poor prognosis 34 . GRN was also secreted from the bone marrow cells and supported the stromal activation and tumor growth in mouse breast cancer models 35 . GRN-secreting monocytes play a key role in the liver metastasis of pancreatic ductal adenocarcinoma 36 . AMBP plays a role in the regulation of inflammation and was reported to be a biomarker of prostate cancer and bladder cancer by proteomic analysis of urine 29,37 . High levels of serum AMBP predict the poor response of gastric cancer patients treated with chemotherapy 38 . CHMP4A, CHMP4C, and CHMP2B belong to the chromatin-modifying protein/charged multivesicular body protein family and are components of ESCRT-III (endosomal sorting complex required for transport III) involved in the formation of endocytic multivesicular bodies 39 . CHMP4A expression is associated with the recurrent ovarian cancer 40 , and CHMP4C plays roles in radiation resistance in non-small cell lung cancer 41 . GRN, AMBP, CHMP4A, CHMP4C, and CHMP2 may play important roles in aggressive prostate cancer and may be potential targets of treatment.
This study has several limitations. First, the sample size is small, and the issue of multiple tests for analyzing the iTRAQ results exists. Further large-scale studies are warranted. Second, there are several hurdles to applying these findings to clinical settings. The development of simpler and easier measurements is mandatory. The isolation of EVs by multiple ultracentrifugation is not suitable in the clinical setting because the isolation process takes more than 4 hours. Recently, several one-step isolation kits for EVs have been developed 42 whose use might circumvent these issues. Proteins located on the membrane of EVs can be measured by these new methods in a high-throughput manner. GO showed that FABP5 is located in the cytoplasm, and FABP5 could also be located inside the EVs. To measure the protein inside the EVs, ExoScreen, which uses two types of antibodies against the proteins located on the membrane of EVs, could not be applied 22 . A patent entitled "Biomarker compositions and methods (WO 2014082083 A1)" describes a technique to detect cancer-associated EVs in bodily fluids by staining lipid layers of EVs, and various cancer-associated antigens including FABP5 were described as potential microvesicle-associated antigens. This method could be a possible technique to detect EVs without ultracentrifugation in clinical settings. Recently, mass-spectrometry analysis has been brought to clinical tests in the hospital. After isolation of the EVs from urine with the one-step kit, biomarkers could be quantified by mass-spectrometry. Further development of mass spectrometry technology might solve these problems. Third, the biological meanings of high amounts of FABP5 in urinary EVs from patients with high GS should be studied.
In conclusion, we applied the proteomic analysis to discover biomarkers in EVs in urine collected after prostate massage. FABP5 in urinary EVs could be a potential biomarker of high GS prostate cancer. Additional large-scale studies are warranted to confirm this finding.

Material and Methods
Sample Collection. Urine samples were collected in Osaka University Hospital. Approval was obtained from the Osaka University Graduate School of Medicine Institutional Review Board before initiating the study, and all patients gave written informed consent. All methods were performed in accordance with the relevant guidelines and regulations. Initial voided urine (20-100 mL) was prospectively collected from patients before prostate biopsy, immediately following DRE. In all cases, the DRE was performed as 3 finger strokes per prostate lobe. Voided urine samples were kept at 4 °C for up to 6 h prior to preparing aliquots of the urine samples and then centrifuged at 2000 × g for 30 min. After removing the pellets, supernatants of urine were stored at −80 °C until analysis.

Patient Classification. We determined Gleason grades from the reports of the pathologists in Osaka
University Hospital based on classifications decided at the 2005 ISUP Consensus Conference. The "negative" group comprised the patients who received prostate biopsy due to elevated PSA levels and were diagnosed pathologically as negative. The "GS6", "GS7", and "GS8-9" groups included the patients diagnosed pathologically as having prostate cancer with GS 6, GS7, and GS 8 or 9, respectively. The "PCa" group was defined as patients with prostate cancer of any GS. In the verification cohort, low-risk prostate cancer was defined as GS6 prostate cancer, and high-risk prostate cancer was defined as GS7, GS8, or GS9 prostate cancer. In the discovery cohort for iTRAQ analysis, 18 samples (negative: n = 6; GS 6: n = 6; GS 8-9: n = 6) were analyzed. In the verification cohort for SRM/MRM analysis, 29 samples (negative: n = 11, the low risk PCa: n = 5, the high risk PCa: n = 13) were analyzed. EV Isolation. Urine samples were centrifuged at 17,000 × g for 30 min to remove debris and then ultracentrifuged at 100,000 × g for 90 min. The supernatant was removed, and the pellets were washed with PBS and ultracentrifuged at 100,000 × g for 90 min. To remove the Tamm-Horsfall protein, the pellets were incubated with DTT for 30 min and ultracentrifuged at 100,000 × g for 90 min. The pellets containing EVs were suspended in 20 μ L of PBS and frozen at −80 °C. EVs were solubilized with MPEX PTS reagent solution (GL Science, Tokyo, Japan) at 95 °C for 5 min followed by sonication for 5 min using a Bioruptor sonicator (Cosmo Bio, Tokyo, Japan). After centrifugation at 100,000 × g for 30 min at 4 °C, the supernatant was obtained as an EV fraction extract and quantified using Pierce Microplate BCA Protein Assay Kit-Reducing Agent Compatible (Thermo Scientific, Waltham, MA, USA).

SDS-PAGE and Western Blot Analysis.
Protein samples were electrophoretically separated on a precast Novex 4-12% Bis-Tris NuPAGE gel using MOPS running buffer (Invitrogen, Carlsbad, CA) according to the manufacturer's instructions. The gels were stained with SYPRO Ruby Protein Gel Stain to confirm the sample quality. Other gels were transferred to polyvinylidene difluoride membranes (Millipore, Bedford, MA). Membranes were immunoblotted with mouse monoclonal anti-human CD9 antibodies (12A12; Shionogi, Osaka, Japan) or rabbit monoclonal anti-human FABP5 antibody (D1AA7T; Cell Signaling Technology, Danvers, MA, USA) followed by horseradish peroxidase-conjugated secondary antibodies and developed with a Super Signal West Dura Extended Duration Substrate kit (Thermo Fisher Scientific).
SCIeNtIfIC REPORTS | 7:42961 | DOI: 10.1038/srep42961 Transmission Electron Microscopy (TEM). Samples of EVs (10 μ g) were adsorbed onto a formvar/ carbon-coated nickel grid for 1 h. EVs were fixed with 2% paraformaldehyde and then reacted with the first antibody (anti-CD9 antibody). Immunoreactive EVs were visualized with the second antibody preabsorbed with 20-nm gold particles (anti-mouse IgG antibody). The samples were negatively stained with 2% aqueous uranyl acetate for 15 min and observed with a JEM-1400Plus transmission electron microscope (JEOL Ltd., Tokyo, Japan).
iTRAQ Labeling. iTRAQ labeling was performed as previously described 21 . The reference pool was arranged by mixing an equal amount (40 μ g) of 18 EV protein extracts. The EV fraction extract (15 μ g) for iTRAQ labeling was reduced with 5 mM DTT for 30 min and then alkylated with 50 mM iodoacetamide for 30 min at room temperature after the addition of bovine serum albumin as the internal standard. The sample was digested with 1% trypsin overnight at 37 °C. The phase-transfer surfactants method was used to remove the surfactants 21 . The tryptic digest sample was desalted using C18 stage tips and then suspended in 30 μ L of iTRAQ dissolution buffer and labeled with 4-plex iTRAQ reagents (Applied Biosystems, Foster City, CA) for 1 hr at room temperature. The tryptic digests of the reference pool, negative group (n = 6), prostate cancer with GS 6 (GS 6 group, n = 6), and prostate cancer with GS 8 or 9 (GS 8-9 group, n = 6)) were labeled with 4-plex iTRAQ reagents 114, 115, 116, and 117 ( Table 1). The labeled samples were then pooled and desalted using C18 stage tips. iTRAQ-labeled peptides were divided into 58 fractions using an HPLC system (Shimadzu Prominence UFLC). We monitored the concentrations of these fractions by UV spectroscopy and then combined low-concentration fractions, which resulted in 17 fractions. The 115:114, 116:114, and 117:114 ratios indicated the relative abundance of proteins in the negative group, GS 6 group, and GS 8-9 group, respectively, relative to the common reference pool. SRM/MRM Analysis. We used SRM/MRM to confirm and further verify the biomarker candidates obtained from iTRAQ. SRM/MRM was performed as previously described 10,11 . A total of 27 EVs were isolated from urine samples from men with negative biopsy (n = 11), men with low-risk prostate cancer (n = 5), and men with high-risk prostate cancer (n = 13) ( Table 1). Stable synthetic isotope-labeled peptides (SI peptides) with a C-terminal 15N-and 13C-labeled arginine or lysine residue (isotopic purity > 99%) were purchased from Greiner Bio One (Frickenhausen, Germany) (crude purity). The peptide sequence was selected from the unique peptide sequences identified in the iTRAQ experiments. Among the 17 proteins with iTRAQ ratios of cancer/ negative biopsy of > 1.5 (p-value < 0.05), six proteins were excluded because appropriate SI peptides could not be designed. The correction for multiple tests was not performed. The SI peptide mixture was analyzed by the above-mentioned LCs-MS/MS method using LTQ Orbitrap-XL to acquire MS data. A preliminary SRM/MRM transition list for SI peptides was created from the MS data acquired using Pinpoint ver.1.0 (Thermo Fisher Scientific, Bremen, Germany). An EV fraction extract (2 μ g) prepared from validation sets of urine samples was reduced with DTT, alkylated with iodoacetamide, and then digested as described above for quantitation using SRM/MRM. The digested peptide was analyzed using the above-described optimal-timed SRM/MRM method with a TSQ-Vantage triple quadruple mass spectrometer (Thermo Fisher Scientific, Bremen, Germany). The parameters used for SRM/MRM analysis were set as follows; scan width of 0.002 m/z, Q1 and Q3 resolution of 0.7 fwhm, cycle time of 2.5 s, and a gas pressure of 1.8 mTorr. Collision energy (CE) was optimized for each transition by performing a test run of the SI-peptide mixture. The SI peptide mixture was added to the trypsin-digested sample, and the area ratio of the endogenous peptide to the SI peptide was calculated using the transition peak area measured with Pinpoint software. Because the SI peptides used were of crude purity, the quantitated values of the endogenous peptide are relative amounts between samples and not absolute amounts. The amount of each SI peptide added to each sample was adjusted to be similar to the endogenous peptide estimated by the peak area obtained from preliminary SRM/MRM of the sample mixture. The average of these ratios of more than two transitions was first calculated, and the average ratio of two technical replicates of an individual sample was then SCIeNtIfIC REPORTS | 7:42961 | DOI: 10.1038/srep42961 determined as the relative quantitative value of the target peptide. The signal-to-noise ratio was identified using Pinpoint software. CD9 proteins were also measured to normalize EVs in urine.

Immunohistochemistry.
Immunostaining was performed as previously reported 43 . Briefly, sections were deparaffinized using xylene and alcohol and incubated with 0.3% H 2 O 2 to block endogenous peroxidase activity. Before immunostaining, the antigen was retrieved by immersing the sections in 10 mmol/L citrate buffer (pH 6.0) and placing them in steam above boiling water for 20 min. Immunohistochemical staining for FABP5 was carried out with anti-endoglin antibody (D1A7T, CST 1:600) using an EnVision + Detection System (DAKO, Glostrup, Denmark) according to the manufacturer's instructions. Primary antibodies were incubated overnight at 4 °C and counterstained with hematoxylin. Data Analysis. Identified proteins were analyzed using a DAVID for GO cellular component and biological process annotation 44 . All statistical analyses were carried out using SPSS version 11.0.1 (SPSS, Chicago, IL, USA) and R version 2.13.0 with RcmdrPlugin. EZR package (Saitama Medical Center, Jichi Medical University), which is a graphical user interface for R (The R Foundation for Statistical Computing). More precisely, it is a modified version of R commander (version 1.6-3) designed to add statistical functions frequently used in biostatistics. Mann-Whitney tests were used to analyze the difference between two categories. Stepwise associations between GS and FABP5 levels were analyzed using the Jonckheere-Terpstra test. Significant factors predicting a GS ≥ 7 were identified by logistic regression analysis, and variables entered into the model were patient age and levels of PSA, PSA density, and FABP5.