Immunoseroproteomic Profiling in African American Men with Prostate Cancer: Evidence for an Autoantibody Response to Glycolysis and Plasminogen-Associated Proteins*

African American (AA) men suffer from a disproportionately high incidence and mortality of prostate cancer (PCa) compared with other racial/ethnic groups. Despite these disparities, African American men are underrepresented in clinical trials and in studies on PCa biology and biomarker discovery. We used immunoseroproteomics to profile antitumor autoantibody responses in AA and European American (EA) men with PCa, and explored differences in these responses. This minimally invasive approach detects autoantibodies to tumor-associated antigens that could serve as clinical biomarkers and immunotherapeutic agents. Sera from AA and EA men with PCa were probed by immunoblotting against PC3 cell proteins, with AA sera showing stronger immunoreactivity. Mass spectrometry analysis of immunoreactive protein spots revealed that several AA sera contained autoantibodies to a number of proteins associated with both the glycolysis and plasminogen pathways, particularly to alpha-enolase (ENO1). The proteomic data is deposited in ProteomeXchange with identifier PXD003968. Analysis of sera from 340 racially diverse men by enzyme-linked immunosorbent assays (ELISA) showed higher frequency of anti-ENO1 autoantibodies in PCa sera compared with control sera. We observed differences between AA-PCa and EA-PCa patients in their immunoreactivity against ENO1. Although EA-PCa sera reacted with higher frequency against purified ENO1 in ELISA and recognized by immunoblotting the endogenous cellular ENO1 across a panel of prostate cell lines, AA-PCa sera reacted weakly against this protein by ELISA but recognized it by immunoblotting preferentially in metastatic cell lines. These race-related differences in immunoreactivity to ENO1 could not be accounted by differential autoantibody recognition of phosphoepitopes within this antigen. Proteomic analysis revealed differences in the posttranslational modification profiles of ENO1 variants differentially recognized by AA-PCa and EA-PCa sera. These intriguing results suggest the possibility of race-related differences in the antitumor autoantibody response in PCa, and have implications for defining novel biological determinants of PCa health disparities.

terminants influencing these disparities (11,12). Reducing these disparities requires a better understanding of PCa biology in different racial contexts, more effective therapeutic regimens for the different stages of the disease, and the availability of novel biomarkers that could be detected in minimally invasive tests for the early diagnosis and management of the disease, particularly in AA men. Prostate specific antigen (PSA) is currently the only widely available bloodbased biomarker for early PCa detection. Although this minimally invasive test offers high sensitivity, its limited specificity often leads to unnecessary biopsies (13). Furthermore, younger AA men, obese AA men, or male smokers may have lower PSA levels despite the presence of aggressive tumors (14 -19). This highlights the urgent need to identify novel biomarkers that complement PSA for early PCa detection with high sensitivity and specificity in diverse populations.
Candidate body fluid-based PCa biomarkers identified to date include autoantibodies to tumor-associated antigens (TAAs), urine-derived fusion proteins, and protein/RNA cargo inside exosomes and other extracellular vesicles (20 -24). Anti-TAAs autoantibodies are promising PCa biomarkers because of their long-term stability in serum, appearance in circulation years before cancer diagnosis, and their status as "reporters" of molecular events associated with PCa progression (25)(26)(27). When profiled against carefully designed TAA arrays these antibodies can distinguish between cancer and noncancer patient groups (25, 28 -30). In addition, these autoantibodies can serve as tools in the identification and molecular characterization of oncoproteins that could serve as potential therapeutic targets (25,26). Examples of TAAs that elicit an autoantibody response in PCa patients, and that have been thoroughly validated as oncoproteins relevant to PCa include glucose-regulated protein 78 (GRP78), cyclin B1, and lens epithelium derived growth factor p75 (LEDGF/p75) (25, 28 -32).
Our knowledge of body fluid-based PCa biomarkers that are differentially associated with AA men compared with European American (EA) men is very limited. Recently, circulating levels of the stem cell marker BMI1 were found in both AA and EA men with PCa, and although BMI1 autoantibodies were found at a higher frequency in PCa patients compared with men without PCa, their association with race/ethnicity was not reported (23,33). The production by PCa patients of cyclin B1 autoantibodies with different FcyRIIIa chains that impair antibody affinity for cyclin B1 was reported to be more associated with AA than EA (32). In a recent study, our team reported differences in serum exosomal protein profiles between AA and EA men with PCa (34). One of these proteins, survivin, a well-established oncoprotein and TAA targeted by autoantibodies in PCa and other cancers, is released by prostate tumors into exosome microparticles at levels that are higher in AA men than in EA men (20). A recent study also demonstrated that AA patients with PCa have increased circulating levels of advanced glycation end products compared with EA patients, underscoring the biomarker potential of these products (35).
Compelling evidence shows that prostate tumors derived from AA and EA men with PCa exhibit race-related differences in their molecular profiles (36 -39), and that in AA men these tumors are localized in an area of the prostate that is rarely surveyed during routine biopsies (40). In addition, race-related differences in the immunobiology of PCa have been reported, with AA prostate tumors displaying increased expression of genes associated with immune responses and inflammation compared with EA prostate tumors (36,38). Yet, most studies on autoantibody profiling and TAA identification in PCa have neither included AA populations nor reported racial/ethnic differences in autoantibody signatures. Given that a growing number of TAAs relevant to PCa have been linked to various key oncogenic pathways operating in this malignancy (22-25, 27-30, 41), it is therefore critical to determine if there are differences in anti-TAA autoantibody responses between AA and EA patients with PCa that could reflect racial differences in tumor immunobiology. Dissecting these differences is essential for designing novel biomarker combinations for PCa detection and management that could complement PSA testing and be tailored to AA men or men at high risk of developing PCa, and for identifying novel biological determinants underlying PCa health disparities.
We used a well-established immunoproteomic approach known as serological proteome analysis (SERPA) to profile autoantibody responses in the sera of AA and EA PCa patients to identify their target substrates (42,43). Our results indicate that several glycolytic enzymes that also participate in plasminogen binding, particularly alpha-enolase (ENO1), are targets of autoantibodies in AA patients with PCa. ENO1 is a metabolic enzyme of ϳ50 kDa that catalyzes in the cytosol the dehydration of 2-phospho-D-glycerate to phosphoenolpyruvate in glycolysis and the reverse reaction in gluconeogenesis, thus playing a role in the anaerobic glycolysis required for the increased rate of cancer cell proliferation (44,45). Autoantibodies to glycolytic enzymes other than glyceraldehyde-3-phosphate-dehydrogenase (GAPDH) have not been reported in PCa patients even though glycolysis genes are overexpressed in various cancers, including PCa (27,46). There is, however, evidence for serum anti-ENO1 autoantibodies in patients with pancreatic cancer (47). This study implicates ENO1 and other enzymes associated with the glycolysis and plasminogen pathways in eliciting autoantibodies in PCa patients, and provides evidence for differences in anti-ENO1 humoral response between AA and EA patients.

EXPERIMENTAL PROCEDURES
Study Subjects and Sample Collection-Sera from patients with PCa (n ϭ 59, self-identified AA, median age 71 yr (average 70); n ϭ 98, self-identified EA, median age 66 yr (average 67)), were obtained from Loma Linda University (LLU) Cancer Center Biospecimen Laboratory, Bioserve (Beltsville, MD), LLU Medical Center, and the serum bank in the LLU Center for Health Disparities and Molecular Medicine.
Normal human sera (NHS, n ϭ 78) were obtained from the serum bank in the Cancer Autoimmunity Research Laboratory of The University of Texas at El Paso (UTEP), and used as controls in this study. These non-PCa sera were collected during annual health examinations in adults who had no obvious evidence of malignancy, no autoimmune disease and have been used in our previous studies on PCa autoantibody profiling as NHS controls (28). We collected additional non-PCa sera (n ϭ 105) during community Black Men Health Fairs held in Riverside, California, and Brooklyn, New York, that focused on disease prevention and health promotion, with emphasis on PCa prevention, education, and research. These fairs were organized under Project C.H.A.N.G.E. (Changing Health in Adults with New and Great Experiences), a LLU community outreach initiative comprised of a transdisciplinary team of scientists and health care professionals. Male participants in these fairs volunteered to donate blood for research and completed a comprehensive socio-demographic, lifestyle, health history, and stress questionnaire (140 questions) with validated instruments. Participants gave their written consent allowing research to be conducted on their blood sample donations and their responses to the questionnaire. Most participants in these two fairs self-identified as Black or AA and did not have personal history of PCa. Licensed nurses and phlebotomists collected the blood samples on red top vials with clotting Z-factor (Vacuette). Serum samples were obtained by incubating the vials for 30 min followed by centrifugation at 5000 rpm for 7 min, and separating the serum supernatants into aliquots that were stored at Ϫ80°C. We selected sera from 105 AA male donors who participated in the fairs and who had no personal or family history of PCa, other cancers, chronic infectious diseases, or autoimmune diseases. These sera also served as control, non-PCa sera. Serum samples from subjects for which we did not have racial/ethnic data were excluded from the study. Subject recruitment and blood sample collection were conducted after informed consent, as approved by LLU and UTEP Institutional Review Boards.
One-dimensional Polyacrylamide Gel Electrophoresis and Western Blotting-Total proteins from prostate cell lysates were prepared and separated by one-dimensional SDS-PAGE (NuPage 4 -12%, Thermo-Fisher Scientific), followed by transfer to polyvinyl difluoride (PVDF) membranes (EMD Millipore, Billerica, MA) in transfer buffer (25 mM Tris, 192 mM glycine, 10% methanol, 0.1% SDS). Membranes were blocked with 5% dry milk in Tris Buffer Saline with Tween-20 (TBS-T) (20 mM Tris-HCl, pH 7.6, 140 mM NaCl, Tween-20) for 1 h, and incubated overnight with commercially available primary antibodies or for 2 h with human sera (1:100). After several washes with TBS-T, membranes were incubated with corresponding HRP-conjugated secondary antibodies for 45 min and then washed again several times with TBS-T. Immunoreactive protein bands were detected by enhanced chemiluminescence (ECL, ThermoFisher). In some experiments, the PVDF membranes were cut into individual lane strips that were then incubated with individual patient sera. Membrane strips were carefully aligned next to each other after the final washes with TBS-T. Autoantibody reactivity to protein bands was detected by ECL on X-ray film.
Serological Proteomic Analysis (SERPA)-For SERPA, total proteins (0.5 mg) from PCa cells separated by 2DE were transferred to PVDF membranes in transfer buffer at 25 volts for 90 min. Membranes were then stained with a PVDF-compatible protein reversible stain (ThermoFisher), and several reference circles were drawn around prominent protein bands/spots on the stained PVDF membranes to serve as orientation and alignment marks. These marks were essential for accurate digital overlay of immunoreactive spots in the film of the PVDF membrane (from analysis 2D gel) probed with human sera with corresponding protein spots in the Coomassie-stained reference 2D gel. Molecular weight markers on the sides of the 2D gels also facilitated the alignment. The reference and analysis 2D gels were run in the same apparatus under identical conditions to ensure accurate alignment of protein spots. PVDF membranes containing proteins transferred from the analysis 2D gels were blocked for 1 h with 5% milk in TBS-T and probed with human sera at a dilution of 1:1000 or 1:500 for 2 h. After extensive washing, membranes were incubated with secondary goat anti-human IgG/IgM/IgA at 1:5000 for 45 min. The immunoreactive protein spots were detected by ECL.
Sample Preparation for Mass Spectrometry-Digital images from the reference 2D gel (Coomassie Blue-stained) and analysis 2D gel (immunoblot film) were acquired using a Licor-Odyssey scanner and then overlaid with Adobe Photoshop 7.0 software using the reference circle marks drawn around prominent protein spots in the PVDF membranes for accurate alignment. A minimum of 90% overlap of the 10 reference marks from the film and the corresponding proteins in Coomassie-stained reference gel was accepted to properly place overlaying figures. Protein spots/bands from the 2D reference gels that corresponded to immunoreactive spots from the films were excised and destained (50% methanol, 5% acetic acid) overnight. Next, the excised protein samples were incubated in 100 mM ammonium bicarbonate with 10 mM DTT at 56°C for 30 min, and then in 100 mM ammonium bicarbonate with 250 mM iodoacetamide for 30 min at room temperature in the dark. The samples were then treated with Trypsin (20 ng/l, Promega, Madison, WI), a proteolytic enzyme that cleaves at the C-terminal amide of lysine and arginine residues, on ice for 30 min and then overnight with 100 mM CaCl 2 at 37°C. The next day, peptide extractions were collected using aqueous solutions and organic acetonitrile (ACN) solutions with 5% formic acid (FA) and then dried using a Savant™ SpeedVac™ concentrator. Samples were immediately reconstituted for desalination using C18 ZipTips (P100, ThermoFisher) according to manufacturer's protocol.
Identification of Candidate TAA by Mass Spectrometry-For mass spectrometry (MS) analysis, an Easy-nLC liquid chromatography system (ThermoFisher) with an autosampler was attached to an LTQ-Orbitrap Velos Pro mass spectrometer (ThermoFisher). Injection of 10 l of peptide samples in 0.1% FA was then passed through a 2-cm C18 pre-column (100 m diameter, 5-micron particle size, Thermo-Fisher). Samples were then eluted and separated on a 10-cm analytic C18 column (75 m diameter, 3-micron particle size, ThermoFisher) with a 2 h linear gradient (5% ACN to 30% ACN in 0.1% FA). Collision-induced disassociation was used to fragment the top 10 most abundant ions and the MS/MS spectra were collected between 250 and 1500 m/z following the parent full scan mass spectrum collected at 60,000 resolution for 2 h. The resulting MS/MS spectra were uploaded and analyzed in Proteome Discoverer 1.4 using associative SEQUEST HT search wizard against the comprehensive Uni-Prot protein sequence database (human, version released March 2013, 89430 sequence entries). The following search parameters were used: two missed trypsin cleavage allowed, dynamic oxidation on methionine, deamidation on asparagine and glutamine, and static carbamidomethylation on cysteine. Mass tolerance for precursor ions was 10 ppm and the mass tolerance for fragment ions was 0.8 Da. High confidence peptide filters with a target of 0.01 false discovery rate (FDR) and an overall score of 100 or higher was used to determine candidate PCa TAA hits. The FDR application counts the matches that pass a given set of filter thresholds from the decoy database and from the nondecoy database. It counts only the top match per spectrum, assuming that for any given spectrum only one peptide can be the correct match. All of the samples' results had an actual FDR below the 0.01 target threshold. The final filtering process of the top scores of candidate TAA hits included identification from multiple sera, the correct corresponding molecular weight and PI in the gel/film, percent coverage, number of unique peptides, and the total number of proteins identified (supplemental Table S1).
For optimization of the SERPA analysis we used commercial antibodies against beta-actin, tubulin, and ENO1, and the corresponding MS/MS spectra of immunoreactive spots confirmed the identity of these proteins. Keratin contaminants were removed from consider-ation as candidate TAAs. For MS/MS analysis of ENO1 post-translational modifications (PTMs), parameters included dynamic phosphorylation on serine, tyrosine, and threonine; acetylation on histidine and lysine; methylation on aspartate, glutamate, histidine, lysine, asparagine, glutamine, and arginine; citrullination on arginine; and glycosyl on proline, asparagine, arginine, serine, threonine, and tyrosine. Supplemental Table S2 lists all the ENO1 peptides with associative PTMs identified from purified ENO1, and the PC3 and PC3-DR cell lines. The MS proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE partner repository with data set identifier PXD003968 (49). All of the raw data files were uploaded and linked to both spectra and result files where the candidate PCa TAA was listed in the experimental factor field. Briefly, msf spectra were converted to mgf files using Proteome Discoverer 1.4 search input and also converted to mzid files with a relaxed peptide filter criteria using ProCon PRIDE for visualization compatibility with ProteomeXchange PRIDE database (50). Files stored in the online repository can be accessed with username "reviewer36284@ebi.ac.uk" and password TA5tDk9M.
Enzyme-linked Immunosorbent Assay (ELISA)-Purified ENO1 used in previous ELISA studies to measure the autoantibody response to this protein in patients with liver fibrosis (51) was diluted in phosphate-buffered saline (PBS) to a final concentration of 0.5 g/ml and coated into 96 well Immunolon2 microtiter plates (ThermoFisher Scientific), which were then incubated overnight at 4°C. Human sera diluted at 1:100 were then added to the antigen-coated wells and incubated for 2 h at room temperature. After several washes with PBS containing 0.05% Tween 20 (PBS-T), HRP-conjugated goat antihuman IgG (1:4000) (Santa Cruz Biotechnology) was added to the wells and incubated for 90 min at room temperature. After additional washes with PBS-T, the substrate 2,2Ј-azino-bis (3-ethyl-benzothiazoline-6-sulfonic acid) diammonium salt (ABTS, Sigma) was used as the detecting agent. The optical density (OD) value of each well was read at 405 nm in a plate reader, and the cut-off value of 0.174 for determining a positive reaction was designated as the mean absorbance of the non-PCa controls plus two standard deviations (meanϩ2SD). Each serum sample was tested in duplicate.
Western Blotting with Phosphatase Treatment for Dephosphorylation Analysis-Total proteins from PC3 cell lysates were prepared and separated on 1D gels followed by transfer to PVDF membranes. The membranes were cut into strips, blocked for 1 h with 5% BSA in tris buffer saline (20 mM Tris-HCl, pH 7.6, 140 mM NaCl), and then treated with lambda -phosphatase (Sigma, Cat# P9614) overnight at 30°C according to the manufacturer's protocol. Briefly, the strips were incubated in phosphatase buffer (Sigma, Cat# L9288) and 2 mM MnCl2 in 5% BSA/TBS with or without 600 U/ml of -phosphatase. The strips are then washed several times with TBS-T (0.1% Tween 20) and blocked for 1 h with 5% milk or BSA in TBS-T. Next, the strips were incubated with mouse anti-AKT (Cell Signaling, Cat# 2920S) or rabbit anti-phosphorylated-AKT (Cell Signaling, Cat# 9271) antibodies diluted 1:1000 in 5% BSA/TBS-T overnight at 4°C, or with human sera diluted 1:100 in 5% milk/TBS-T for 2 h. The strips were then washed with TBS-T, incubated with secondary antibodies for 45 min, washed again with TBS-T, and then exposed to X-ray films using ECL as previously stated.
Experimental Design and Statistical Rationale-A test cohort of 89 PCa patient sera was initially probed against metastatic PC3 total cell lysates using WB. Complementary numbers of PCa sera from AA and EA men were used and analyzed under identical conditions for comparison. Candidate TAAs reported were based on SERPA analysis from multiple strongly immunoreactive PCa sera. ELISA studies measuring anti-ENO1 autoantibody frequency were performed in duplicate using a larger validation cohort of 340 sera including 183 non-PCa sera and 157 PCa sera. The Chi square test for independence was used to determine p values for statistical significance and distinguish anti-ENO1 autoantibody frequency between the PCa and non-PCa cohorts. A p value of 0.05 or less was accepted as a determinant for statistical significance.

Screening of PCa Sera for the Presence of Autoantibodies
to Potential TAA-Our immunoproteomic approach for the profiling of serum autoantibodies targeting candidate TAAs in AA and EA men with PCa is illustrated in supplemental Fig.  S1. In initial experiments, we screened by WB 89 sera from AA and EA PCa patients for the presence of anti-TAA autoantibodies against total protein lysates from metastatic PC3 cells. PC3 proteins were separated by SDS-PAGE and transferred to PVDF membranes which were then cut into individual lane strips to be incubated with individual patient sera and secondary anti-human IgG/IgM/IgA antibody. The strips were aligned next to each other, and immunoreactivity to protein bands was detected by ECL on the same film. A comparison of randomly selected sera from AA-PCa patients (n ϭ 24) and EA-PCa patients (n ϭ 27), under identical experimental conditions, revealed that the AA-PCa sera had stronger immunoreactivity to PC3 proteins than the EA sera on the same film exposed to ECL reagent for 1 min (Fig. 1A, 1B). After increasing the exposure time to 3 mins, immunoreactive bands were visible for some EA-PCa sera (Fig. 1D), whereas most of the AA-PCa sera showed dramatically increased immunoreactivity that darkened the film (Fig. 1C). Several AA-PCa sera showed common immunoreactivity around the 50 kDa region (16 out of 24, 67%) (Fig. 1E), whereas only one of the 27 randomly selected EA-PCa sera (EA1-PCa, 3.7%) showed relatively strong immunoreactivity in this region after film FIG. 1. Screening of PCa patient sera for autoantibodies to candidate tumor associated proteins by Western blotting. Each number corresponds to a different patient serum probed against whole PC3 cell lysates in PVDF membrane strips. Serum antibody reactivity against individual proteins for each patient was detected using HRP-labeled anti-human IgG/IgM/IgA secondary antibody and enhanced chemiluminescence (ECL) on film. A, One-minute exposure of membrane strips to film using sera from AA men with PCa. B, One-minute exposure of membrane strips to film using sera from EA men with PCa. C, Exposure increased to 3 mins showing overexposed AA serum immunoreactivity. D, Exposure increased to 3 mins to visualize EA serum immunoreactivity. E, Western blot zoomed in to highlight common 50 kDa immunoreactivity of several AA-PCa sera. All membrane strips were simultaneously probed with sera and exposed to ECL under identical conditions in the same experiment. *AA-PCa sera with 50 kDa immunoreactivity were identified to have anti-ENO1 antibodies using SERPA.
overexposure. Other common bands of different molecular weights were also present in multiple sera including reactivity below 37 kDa, and bands around 40, 55, 60, and 65-70 kDa. These differences in serum immunoreactivity between AA-PCa and EA-PCa patients in the test cohort were independent of age or tumor stage.
Identification of Candidate TAAs Associated with Glycolysis and Plasminogen Pathways-We selected representative AA-PCa sera with strong immunoreactivity including those reacting with a protein band around 50 kDa (Fig. 1) and examined them by SERPA. Images of immunoblots of 2D analysis gels exposed on films were overlaid and digitally matched with images of the corresponding Coomassie Blue-stained reference 2D gels (Fig. 2). Protein spots in stained 2D gels corresponding to the immunoreactive spots were processed for LTQ-Orbitrap-MS and the resulting MS/MS spectra were analyzed by SEQUEST software using the UniProt database. The filtering process for top scores of protein hits in the excised immunoreactive spots was based on molecular weight, isoelectric point, percent coverage, number of unique peptides, and total number of proteins identified for each spot. For positive controls we used commercial antibodies against known proteins, and the corresponding MS/MS spectra of 2DE spots reacting with these antibodies confirmed the identity of their target proteins (data not shown). ENO1, an enzyme that participates in both glycolysis and plasminogen pathways (44,52), was identified as the top hit from the 50 kDa spot recognized by each of six selected AA-PCa sera; representative mass spectra of two ENO1 peptides from the excised 50 kDa spot shown in Fig. 3.
Interestingly, SERPA analysis of the other strongly immunoreactive spots (outside the 50 kDa region) recognized by these and other highly reactive PCa sera (n ϭ 14, 11 AA-PCa and 3 EA-PCa) revealed additional candidate TAAs involved in glycolysis and plasminogen pathways (Table I). A total of 20 protein hits were identified from the analysis of 2DE spots recognized by these 14 highly reactive PCa sera. In addition to ENO1, these protein hits included key glycolytic enzymes such as glyceraldehyde-3-phosphate dehydrogenase (GAPDH), fructose bisphosphate aldolase (ALDO), phosphoglycerate kinase (PGK), and lactate dehydrogenase (LDH). The most common serum reactivities detected by WB were against protein bands of ϳ75 kDa and 50 kDa, which were identified by SERPA as GRP78 and ENO1, respectively. Annexin A2, a plasminogen associated protein, was also identified as a target of autoantibodies in our AA-PCa test group. It is interesting to note that ENO1, GAPDH, and GRP78 have also been implicated in plasminogen signaling (Table I). Additional candidate TAAs identified by our SERPA approach have been associated with cell adhesion, stress survival, and cellular metabolism ( Table I). Several of these TAAs have been validated as aberrantly expressed in PCa by immunohistochemistry (IHC) and quantitative polymerase chain reaction (qPCR) ( Table I). FIG. 3. Mass spectrometry chromatograph and associated mass spectra of representative ENO1 peptides identified from excised 2D gel protein spots corresponding to immunoreactive spots recognized by two different sera. A, Chromatograph and select mass spectra of two ENO1 peptides with retention times (RT) that were identified after probing with PCa-AA10 serum. B, Chromatograph and mass spectra of two ENO1 peptides identified after probing with PCa-AA18 serum.

Confirmation of ENO1 as a Candidate
PCa TAA-Of the 14 PCa sera that were screened by 2DE-WB because of their high immunoreactivity against total proteins in PC3 cell lysates (Fig. 1), six were from AA PCa patients (PCa-AA7, PCa-AA9, PCa-AA10, PCa-AA11, PCa-AA18, and PCa-AA26) who had autoantibodies to ENO1 as determined by SERPA. To further confirm the presence of anti-ENO1 autoantibodies in these AA-PCa sera, we compared the immunoreactivity of selected AA sera with that of a commercially available monoclonal antibody raised against recombinant human ENO1 (MoAb-ENO1). This antibody reacted with both 50 kDa ENO1 and its 37 kDa splice variant MBP1 by 1D-WB, and the selected AA-PCa sera showed the same pattern of immunoreactivity as MoAb-ENO1 (Fig. 4A). As additional validation, we observed identical migration patterns for purified ENO1 protein and ENO1 from PC3 cell lysates (PC3-ENO1) in 2DE, with pI of 7.29 in the first dimension and molecular weight of 50 kDa in the second dimension ( Fig. 4B-4E). Alignment of immunoblots of the 2D gels showed exact alignment of the immunoreactive ENO1 protein spots (both from purified ENO1 and PC3-ENO1) recognized by both MoAB-ENO1 and representative PCa-AA10 serum (Fig. 4C-4E).
Higher Frequency of Anti-ENO1 Autoantibodies in PCa Sera Than in Controls-After demonstrating that several sera in our initial PCa patient test cohort contained autoantibodies to ENO1 we then set up a validation ELISA, using purified ENO1 as target substrate (49), to determine the frequency of these autoantibodies in an expanded validation cohort of AA and EA-PCa patients (n ϭ 157), as well as a non-PCa control cohort (n ϭ 183). These cohorts contained all of the sera shown in Fig. 1 plus additional sera acquired at LLU, UTEP, and Black Men Community Health Fairs. As shown in Fig. 5A, sera from PCa patients showed a significantly higher frequency (13/157, 8.3%) and levels of autoantibodies to ENO1 compared with non-PCa sera (5/183, 2.7%) (p Ͻ 0.05), suggesting that they might be cancer-related. Fig. 1) we evaluated by ELISA the anti-ENO1 antibody reactivity in sera from AA-PCa and EA-PCa patients. Unexpectedly, we observed that the frequency of anti-ENO1 autoantibodies was higher in the EA-PCa validation cohort than in the AA-PCa cohort in the ELISA analysis ( Fig. 5B; p Ͻ 0.05). Only 1 of 59 (1.7%) AA-PCa sera in this validation cohort reacted with purified ENO1 by ELISA above the cut-off value (determined by mean of non-PCa controls plus two standard deviations) (Fig. 5B), despite the AA-PCa test cohort having a higher frequency of reactivity against PC3-ENO1 by WB (Fig. 1). By contrast, 12 of 98 EA sera (12.2%) reacted with ENO1 by ELISA (Fig. 5B), despite the relatively low frequency of reactivity of the EA-PCa test cohort sera against PC3-ENO1 by WB (Fig. 1).

Differences in Anti-ENO1 Autoantibody Reactivity Between Sera from EA-PCa Patients and AA-PCa Patients-To further validate our initial observations with the test cohorts (
Given these results, we explored whether selected AA-PCa sera that recognized ENO1 in PC3 cells (PC3-ENO1) by WB would also recognize purified ENO1 by WB, and whether selected EA-PCa sera that reacted against ENO1 by ELISA would also recognize both purified ENO1 and PC3-ENO1 by WB. Interestingly, AA PCa sera AA9, AA10, and AA11, which reacted strongly against PC3-ENO1 by WB (Figs. 1A, 4 -5), but weakly against purified ENO1 by ELISA (OD values 0.073, 0.159, and 0.128, respectively), also reacted weakly with purified ENO1 by WB (Fig. 5C). By contrast, two EA-PCa sera that reacted strongly against purified ENO1 by ELISA, EA100 and EA43 (OD values 0.517 and 0.53, respectively), also reacted strongly with purified ENO1 by WB (Fig. 5C) but moderately with PC3-ENO1 (Fig. 5D). EA-PCa serum EA1, which reacted weakly in ELISA (OD value 0.115) and against PC3- FIG. 4. Alignment of ENO1 immunoreactivity in Western blots and 2D gels. A, Western blot of PC3 cell lysates probed with MoAB-ENO1 and select African-American PCa sera that recognized ENO1 in 2D gels showed immunoreactivity at the same 50 KD molecular weight. All the lanes came from the same blot. Note that the MoAB-ENO1 and the select African-American PCa sera also react with a 37 kDa band corresponding to MBP1, a small splice variant of ENO1. B, Silver-stained 2D gel of purified ENO1 (40 ug). C, 2D Western blot of recombinant ENO1 protein probed with MoAB-ENO1. D, 2D Western blot of PC3 cell lysate probed with MoAB-ENO1. E, 2D Western blot of PC3 cell lysate probed with PCa-AA10 serum (1:500). ENO1 ( Fig. 1D and Fig. 5D), reacted strongly against purified ENO1 in WB (Fig. 5C). Interestingly, serum AA56, the only AA-PCa serum that reacted positive by ELISA (OD value 0.530), reacted weakly by WB with both purified ENO1 and PC3-ENO1 (Fig. 5C, 5D). These results revealed heterogeneity across several patients in their anti-ENO1 serum immunoreactivity, with some intriguing differences between the AA-PCa and EA-PCa sera in their reactivity against purified and cellular ENO1.
The differences of anti-ENO1 serum immunoreactivity suggested the possibility that some AA-PCa and EA-PCa sera may recognize different variants of ENO1. To test this possibility, we determined if the AA-PCa and EA-PCa sera displayed the same pattern of immunoreactivity against cellular ENO1 in a diverse panel of 12 prostate cell lines that included non-PCa and PCa cells, as well as AA-and EA-derived cells. The commercial MoAB-ENO1 antibody (raised against recombinant human ENO1), used as internal control, showed a consistent reactivity against cellular ENO1 across the panel of non-PCa and PCa cell lines (Fig. 6A). Similarly, the EA-PCa sera EA100 and EA43 also consistently recognized ENO1 across the panel of cell lines (Fig. 6A). By contrast, the AA-PCa sera AA10, AA11, AA18 showed a selective pattern of immunoreactivity, reacting weakly with ENO1 in the non-PCa cell lines PrEC, RWPE-1, BRF-55T, and RC-77N/E, but stronger in the AA-derived metastatic cell line MDA-PCa-2b and the EA-derived metastatic cell lines LNCaP, PC3, and DU145 (Fig. 6B). Interestingly, unlike the EA-PCa sera and the MoAB, the immunoreactivity of the AA-PCa sera against ENO1 was dramatically reduced in the docetaxel-resistant cell lines PC3-DR and DU145-DR, suggesting that ENO1 may undergo modifications or alterations in these drug resistant cell lines that negatively influence its antigenicity. Taken together, these differences in anti-ENO1 immunoreactivity suggested that AA-PCa and EA-PCa patients may produce autoantibodies to distinct ENO1 variants, and that the variant recognized by AA-PCa patient autoantibodies could be associated with a more aggressive PCa phenotype.
Characterization of ENO1 Post-translational Modifications-Given the precedent that anti-ENO1 autoantibodies in pancreatic cancer patients specifically recognize phosphorylated ENO1 (47), we evaluated the possibility that anti-ENO1 PCa sera also reacted with phosphorylated ENO1 in PC3 lysates. For these studies, -phosphatase was used to dephosphorylate PC3 proteins after they were transferred to PVDF membranes. The blots with and without dephosphorylation were then incubated with selected AA and EA PCa sera. As internal controls, we used antibodies to total and phosphorylated AKT. As predicted, the anti-phosphorylated AKT antibody, but not the anti-AKT antibody, lost immunoreactivity when PC3 proteins in the membrane were treated with -phosphatase. However, PC3 protein dephosphorylation did not influence the anti-ENO1 immunoreactivity of the AA and EA PCa sera (Fig. 7), suggesting that, unlike in pancreatic cancer patients (47), the anti-ENO1 antibodies in PCa patients do not appear to be directed at phosphoepitopes.
These results suggested that perhaps other types of PTMs may influence the immunoreactivity of PCa patients with ENO1. As an initial step in determining if the differences in immunoreactivity to purified ENO1, PC3-ENO1, and PC3-DR ENO1 observed with AA-PCa and EA-PCa sera could be associated with differential PTMs in this protein, we profiled by MS/MS the ENO1 PTMs in these three contexts. Precedent for these experiments lies on previous observations that three types of PTMs, phosphorylation, methylation, and acetylation, can be identified by MS/MS in pancreatic cancer ENO1, and that a phosphorylated ENO1 variant is targeted by autoantibodies in EA patients with this malignancy (47,53).
ENO1 spots on Coomassie Blue-stained 2D gels (from purified ENO1, PC3 lysates, or PC3-DR lysates) were excised and analyzed for PTMs by LTQ-Orbitrap-MS. Parameters for PTM profiling included phosphorylation on serine, tyrosine, and threonine; acetylation on histidine and lysine; and methylation on aspartate, glutamate, histidine, lysine, asparagine, glutamine, and arginine. We also used MS to look for glycosylation and citrullination, two other types of ENO1-PTMs described in non-PCa contexts (54,55). PTM maps generated for purified ENO1, PC3-ENO1, and PC3-DR-ENO1 revealed unique profiles for the three proteins, with similarities and differences in the modification patterns of specific residues ( Fig. 8 and Table II). There was also a large frequency of acetylated and methylated residues in the PC3-ENO1, with lower PTM frequency in purified ENO1 (Table III). These profiles indicated the presence of PTMs that are specific for PC3-ENO1 and that are not present in PC3-DR-ENO1 or purified ENO1. For instance, there were several phosphorylated residues (e.g. Thr41, Tyr131, Tyr189, Thr321) in PC3-ENO1 that were not present in ENO1 from docetaxel-resistant PC3-DR cells and, viceversa, there were phosphorylated residues in the PC3-DR cells (e.g. Ser62, Thr99, Ser263, Ser268, and Ser419) that were not present in the PC3 cells. Likewise, several ENO1 acetylated lysines (e.g. Lys60, Lys228, Lys256, Lys422) and methylated glutamic acid, aspartic acid, and lysine residues (e.g. Glu45, Asp53, Asp98, Lys233, Asp238, Glu250) were also found solely in PC3-ENO1 and not in PC3-DR-ENO1 or purified ENO1. Representative MS/MS spectra for acetylated, methylated, and phosphorylated ENO1 are shown in supplemental Figs. S2-S4. We also observed differences between these PC3-ENO1 profiles and those published for pancreatic cancer ENO1 (53) (data not shown). These differences in the pattern of ENO1 PTM in different contexts might explain the observed differences in anti-ENO1 reactivity between AA-PCa and EA-PCa sera.

DISCUSSION
Although AA men are at a higher risk of developing aggressive PCa compared with other racial/ethnic groups, this population has been underrepresented in most research studies on antitumor autoantibody profiling in PCa. Given the heterogeneity of PCa tumors in diverse male populations, and the recent reports pointing to racial differences in the immunobiology of prostate tumors (32,36,38), we undertook an effort to profile, using an immunoseroproteomics approach, sera from PCa patients of African and European descent for the presence of autoantibodies to potential TAAs. Our observation that sera from AA-PCa patients showed a stronger immunoreactivity against PC3 cellular proteins by WB than sera from EA-PCa patients under identical experimental conditions is consistent with the notion that prostate tumors from AA and EA men may exhibit different molecular phenotypes that could influence differential antitumor immune responses (32,36,38). This observation also strengthens the critical case for inclusion of men of African descent into PCa biomarker studies. Given the disproportionately high PCa-associated mortality in AA men, there is an urgent need for renewed investment into our understanding of antitumor immune responses in racially diverse PCa populations, and how these responses relate to disease aggressiveness and therapeutic outcomes.
Our immunoseroproteomics profiling led to the identification of several potential TAAs targeted by serum autoantibodies in PCa patients, particularly in AA patients. These TAAs included multiple members of the glycolysis and plasminogen pathways, several of which have already been validated as cancer associated proteins in prostate tumors by IHC or qPCR (27, 31, 46, 56 -65). For instance, PGK1 has been linked to increased PCa progression, bone metastasis, and bone formation (59,66,67). LDH is a late stage glycolysis enzyme that converts pyruvate to lactate and its expression has been linked to PCa tumor progression and cell migration in breast cancer (62,68). Although the overexpression of glycolysis genes in PCa tissue has been reported (46), to our knowledge the present study is the first demonstration that the glycolytic enzymes ALDO, PGK1, ENO1 and LDHA are targeted by autoantibodies in PCa patients. The up-regulation of these enzymes in PCa is consistent with the notion that cancer cells have an increased dependence on glycolysis to produce ATP for rapid proliferation (45). Additional studies are needed to determine the frequencies of autoantibodies to these and other metabolic enzymes in diverse PCa patient cohorts.
Metabolic enzymes may also function in altering the tumor microenvironment (66,68). For instance, there is evidence that glycolytic enzymes can translocate from the mitochondria to the plasma membrane during cytoskeletal remodeling, where they can influence cell motility, extracellular matrix degradation, and plasminogen signaling (44,52,69). For example, ENO1 and GAPDH have been reported to "moonlight" as plasminogen receptors, contributing to the conversion of plasminogen to plasmin and promoting extracellular matrix remodeling with ensuing increased cancer cell migration and metastasis (44,52,70). In glioma tumors, ENO1 is overexpressed to promote migration and tissue invasion (71), whereas its secretion from PCa stromal cells was also found to improve cell migration (65). In another study, a small molecular inhibitor of ENO1 reduced pancreatic cancer cell invasion and migration properties (72), suggesting that this glycolytic enzyme plays a significant role in aggressive disease progression. The overexpression of these metabolic enzymes on the surface of tumors could break immune tolerance, making them targets of humoral responses. Intriguingly, anti-ENO1 autoantibodies may assuage tumor metastasis by interfering with plasminogen binding, thus resulting in reduced tumor growth, migration and invasion (73). This makes ENO1 and other glycolytic enzymes contributing to plasminogen binding attractive candidates for anticancer immunotherapy.
Other TAAs identified in this study have been also shown to be overexpressed in PCa tissue and play a role in microenvironment transformation and cancer migration. For instance, like ENO1, annexin A2 is a plasminogen receptor that can promote tissue invasion, and its incorporation in extracellular exosomes has been associated with increased invasiveness (70,74). There is also evidence that annexin A2 is associated with ENO1 in a Caveolin-1 (Cav-1)-dependent manner, because knockdown of both annexin 2 and Cav-1 decreased ENO1 expression on the cell surface (75). GRP78, a member of the HSP70 family, is a well-documented TAA with increased autoantibody frequency in metastatic PCa patients (31). Recently, GRP78 was also identified as a plasminogen receptor in neuroblastoma, and autoantibodies to this protein have been shown to inhibit NF-kappa B activation in PCa, leading to p53 up-regulation and decreased PCa proliferation (76,77). This provides further evidence for a protective role against PCa of autoantibodies to metabolic enzymes that also act as plasminogen receptors. Autoantibodies to plasminogen have also been found in PCa sera (78), consistent with the notion that the plasminogen system is targeted by the immune system in certain PCa patients. A very intriguing observation in our study was the differential anti-ENO1 autoantibody reactivity between AA-PCa and EA-PCa sera in different platforms. Although several AA-PCa sera reacted strongly with ENO1 in PC3 cells, they reacted weakly with purified ENO1 in ELISA. By contrast, EA-PCa sera that reacted strongly with purified ENO1 in ELISA and WB showed moderate reactivity against PC3-ENO1. Furthermore, when selected AA-PCa and EA-PCa anti-ENO1 sera were probed against the cellular protein in a panel of 12 non-PCa and PCa cell lines, two distinct patterns of ENO1 recognition emerged. EA-PCa sera that had high immunoreactivity to purified ENO1 in ELISA and immunoblots showed a uniform ENO1 recognition pattern across the 12 non-PCa and PCa cell lines. This pattern was also observed with a monoclonal anti-human ENO1 antibody. However, several AA-PCa sera identified through SERPA to contain autoantibodies to PC3-ENO1, and that showed low immunoreactivity against purified ENO1 in ELISA and WB, displayed a selective pattern of ENO1 recognition in the panel of cell lines, with increased reactivity against this protein in the aggressive PCa cell lines MDA, LNCAP, PC3, and DU145 and low reactivity in the non-PCa cell lines and in the docetaxel-resistant variants PC3-DR and DU145-DR. Future studies need to be conducted on a panel of cell lines from other cancer types to observe if racial differences in ENO1 immunoreactivity can also be detected.
Although the differential autoantibody response to ENO1 in different platforms could potentially be explained by the recognition of different epitopes (denatured versus conformational) (79), the selective increased reactivity of AA-PCa sera against ENO1 in a particular set of cell lines, compared with the nonselective reactivity of the EA-PCa sera, would be difficult to explain based on differential recognition of autoepitopes. However, although human autoantibodies are typically directed against highly conserved epitopes (26), we cannot categorically rule out that EA-PCa and AA-PCa patients are mounting humoral immune responses that recognize different epitopes within ENO1. Future epitope mapping studies could provide insights into this possibility. The observed differences between the reactivity of EA-PCa and AA-PCa to ENO1 in different cell lines could be more consistent with a dependence on a specific ENO1 PTM profile for autoantibody recognition that is differentially present in metastatic PCa versus non-PCa cells. Precedent for this can be found on the previous observation that phosphorylation of ENO1 at serine 419 triggers autoantibody and T cell responses to this protein in genetically susceptible (HLA-DRB1-8ϩ) Italian patients with pancreatic cancer but not in control subjects (47,51,80). Our ENO1 PTM maps showed phosphorylation of this residue in ENO1 from PC3-DR cells (recognized by our selected EA-PCa sera but not by the selected AA-PCa sera), but this phosphorylation was not found in ENO1 PC3 (recognized by the EA-PCa and AA-PCa sera we tested, as well as the monoclonal antibody). Moreover, our results did not appear to indicate that the PCa sera recognized phosphorylated ENO1 in PC3 cells.
Our proteomic PTM analysis revealed differences between PC3-ENO1, PC3-DR-ENO1, and purified ENO1, in the number and position of methylated and acetylated residues. Acetylation and methylation are two additional types of PTMs that have been previously observed in ENO1 from pancreatic cancer cells, although it is not clear if they influence ENO1 immunogenicity (53). For instance, in our analysis, a couple of acetylated residues and methylated residues that were only found in PC3 cells (e.g. acetylated Lys256 and methylated Asp383) and not in PC3-DR nor in recombinant ENO1 were also identified in the pancreatic cancer PDAC cell line but not in normal pancreatic Panc1 cells (53). Although the role of PTMs in ENO1 function is still unclear, it has been suggested that these modifications play a role in the translocation of this protein from the mitochondria to the plasma membrane (44). There is also evidence in thyroid cancer that the addition of retinoic acid to cancer cell lines dephosphorylated ENO1, reduced ENO1 transcripts, and decreased cellular invasiveness (81). Other types of PTM that could explain the different ENO1 immunoreactivities associated with select AA-PCa and EA-PCa sera are glycolysation and citrullination, two modifications that have been previously described in ENO1 in non-PCa contexts (54,55).
It is plausible that differences in the type and position of PTMs in ENO1 from different PCa cell lines may influence the reactivity of EA-PCa and AA-PCa sera against cellular ENO1, leading to the cell line-dependent selectivity in immunoreactivity observed with select AA-PCa. In the case of drug sensitive and resistant cell lines we observed that the AA-PCa sera, unlike the EA-PCa sera, reacted with ENO1 in the chemosensitive PC3 and DU145 cell lines, but not in the chemoresistant PC3-DR and DU145-DR cell lines. If these striking differences in ENO1 immunoreactivity between AA-PCa and EA-sera were based solely on recognition of different epitopes, then it would be unlikely that the transition from chemosensitivity to chemoresistance would alter an autoantibody response to a conserved epitope in an abundant wild type protein. However, it could be hypothesized that the long-term selection in culture of docetaxel-resistant cells may alter the pattern of PTMs in ENO1, resulting in decreased autoantibody reactivity. This would be consistent with our proteomic data indicating that ENO1 is differentially posttranslationally modified in PC3 versus PC3-DR cells.
Alternatively, it is conceivable that autoantibodies from AA-PCa sera may preferentially recognize ENO1 epitopes containing specific mutations that are expressed at higher rates in the PC3, DU145, MDA, and LNCAP cell lines, but not in the other cell lines. The elicitation of autoantibodies to mutated proteins in PCa is well documented in the case of the p53 tumor suppressor (82). It cannot be ruled out, however, that ENO1 mutations may alter its expression, subcellular localization, and immunogenicity, making it a target of autoantibodies. This would be consistent with the observation that ENO1 mutations can markedly alter its expression and subcellular localization by disrupting its interaction with caveolin-1 lipid rafts (75). Future studies in our laboratories will examine whether a particular type and position of PTM or mutation in PCa ENO1 influences its recognition by autoantibodies from PCa patients, particularly in a race dependent manner. Further analysis of the autoantibody response to ENO1 in PCa could potentially unveil a novel immune determinant that might influence tumor aggressiveness, particularly in AA-PCa men, thus contributing to the observed mortality disparity affecting this population.
The identification and characterization of novel PCa-associated autoantibodies targeting TAAs from the glycolytic and plasminogen pathways could enhance our understanding of the immune system's role during prostate tumorigenesis, as well as provide new tools to fine-tune early PCa diagnosis and management using minimally invasive methods. These autoantibodies could also act as internal biosensors for predicting which men are at high risk of developing aggressive PCa. Although anti-TAA autoantibodies in PCa may have limited diagnostic and prognostic value when used individually, they have shown promise when profiled against carefully designed TAA arrays (25,28,29). These autoantibodies could also provide a platform for the development of novel antibodybased immunotherapies for PCa.
Although active surveillance is currently recommended for men with lower PCa risk, growing evidence suggest that this recommendation should not be broadly applied to men from all racial groups, especially those from high risk populations such as men from African descent (4). It is therefore imperative to increase the inclusion of men from diverse racial/ethnic backgrounds in studies on PCa biology and biomarker discovery. The findings presented in this study, although requiring further investigation to elucidate the molecular basis for the race-related differences in the anti-ENO1 autoantibody response in PCa patients, support the growing evidence for differences in the immunobiology of prostate tumors between AA and EA men, which may translate into racial differences in antitumor immune responses.