Evaluation of ERG Responsive Proteome in Prostate Cancer

Abstract

BACKGROUND

Gene fusion between TMPRSS2 promoter and the ERG proto-oncogene is a major genomic alteration found in over half of prostate cancers (CaP), which leads to aberrant androgen dependent ERG expression. Despite extensive analysis for the biological functions of ERG in CaP, there is no systematic evaluation of the ERG responsive proteome (ERP). ERP has the potential to define new biomarkers and therapeutic targets for prostate tumors stratified by ERG expression.

METHODS

Global proteome analysis was performed by using ERG (+) and ERG (−) CaP cells isolated by ERG immunohistochemistry defined laser capture microdissection and by using TMPRSS2-ERG positive VCaP cells treated with ERG and control siRNA.

RESULTS

We identified 1196 and 2190 unique proteins stratified by ERG status from prostate tumors and VCaP cells, respectively. Comparative analysis of these two proteomes identified 330 concordantly regulated proteins characterizing enrichment of pathways modulating cytoskeletal and actin reorganization, cell migration, protein biosynthesis, and proteasome and ER-associated protein degradation. ERPs unique for ERG (+) tumors reveal enrichment for cell growth and survival pathways while proteasome and redox function pathways were enriched in ERPs unique for ERG (−) tumors. Meta-analysis of ERPs against CaP gene expression data revealed that Myosin VI and Monoamine oxidase A were positively and negatively correlated to ERG expression, respectively.

CONCLUSIONS

This study delineates the global proteome for prostate tumors stratified by ERG expression status. The ERP data confirm the functions of ERG in inhibiting cell differentiation and activating cell growth, and identify potentially novel biomarkers and therapeutic targets.

INTRODUCTION

Carcinoma of prostate is the most frequently diagnosed non-skin cancer in the United States with an estimated 238,590 newly diagnosed cases and 29,720 deaths in 2013 (1). Rapidly increasing understanding of the molecular basis of CaP is providing new insights into the etiology and improved prognosis of the disease (2-4). Prevalent gene rearrangements in CaP involve the fusion promoter region of AR regulated genes (predominantly, serine 2 trans-membrane protease: TMPRSS2) and protein coding sequence of an ETS related gene (primarily ERG). While TMPRSS2-ERG is detected in 40%-65% of patients, SLC45A3 and NDRG1 serve as fusion partners for approximately 10% of the tumors with ERG rearrangements (5-7)

Despite the high prevalence of TMPRSS2-ERG gene fusions detected in CaPs of Western populations, the frequency is lower in African Americans (31%-43%) compared to Caucasian Americans (50-66%), and it is even lower in Asian populations (5-24.4%) (8-10). We have recently reported that ERG frequency is strikingly less in the index tumors of African American patients (28.6%) compared to Caucasian Americans (63.3%), suggesting that the ERG based stratification of CaP may help distinguish the biologic differences of CaP between the ethnic groups (10). Studies comparing ERG (+) and ERG (−) CaP have also suggested the expression of genes unique to ERG (+) or ERG (−) tumors (11,12).

Multiple studies on the ERG regulated transcriptome have investigated the function of ERG in the context of prostate epithelial cells and its effect on tumor cell invasion or prostate epithelial differentiation (13-16). However, the underlying mechanisms of ERG function remain to be better elucidated. Although there have been considerable efforts to characterize the CaP proteome (17-21), a systematic evaluation of ERG responsive proteome (ERP) has not been carried out. Since ERG oncoprotein is a nuclear transcription factor, it is neither an optimal biomarker nor an ideal cancer therapeutic target. The evaluation of ERG Responsive Proteins (ERPs) may identify surrogate biomarkers from secreted or cell surface proteins or druggable targets such as growth factor receptors or kinases in the ERG network. Furthermore, differential expression of proteins in ERG (+) and ERG (−) CaP may delineate the biochemical differences and identify potential biomarkers and therapeutic targets of specific for these two tumor types.

Until recently, the lack of reliable ERG antibodies has restricted the analysis of ERG aberrations in CaP specimens to fluorescence in situ hybridization (FISH) or reverse transcriptase polymerase chain reaction (RTPCR) assays (22,23). We have adopted a novel approach to study the ERG modulated proteome by identifying tumor cells positive or negative for ERG protein expression using ERG-MAb-based immunohistochemistry (IHC) staining of prostate tumor specimens (24), followed by the isolation of cells using laser capture microdissection (LCM) (25). Using ERG siRNA, we also inhibited the expression of the ERG protein in VCaP cells, which enabled us to compare ERG responsive proteome in the presence or absence of ERG.

The application of sensitive and quantitative methods in shotgun proteomics has significantly improved the resolution proteomic of analysis. In this study, we used a unique platform based on capillary isotachophoresis (CITP) and capillary zone electrophoresis (CZE) coupled with electrospray ionization (ESI) linear ion trap tandem mass spectrometry (MS/MS). The combined CITP/CZE-nano-ESI-MS/MS system has been demonstrated to be at least one to two orders of magnitude more sensitive than that found in conventional electrophoresis and column-chromatography based proteome technology, covering a much wider concentration range, necessary for increasing the range of protein profiling (26). This improvement is achieved by the selective analyte enrichment through electrokinetic stacking of CITP, and the excellent resolving power of CZE (27), which results in diluting the major components while concentrating the trace compounds. The on-column transition from CITP to CZE also minimizes additional band broadening with superior analyte resolution.

We defined the differentially expressed ERP both from prostate tumor specimens and from VCaP cell line and revealed a total of 330 overlapping proteins that concordantly respond to ERG expression. Literature-based evaluation for functionally interacting signaling pathways revealed networks regulating multiple cellular functions including AR signaling, protein synthesis and trafficking, and cell growth and migration. By sorting for ERPs that were detected at higher MS ratios in ERG (+) and ERG (−) tumors relative to benign tissues, we sought to distinguish these tumors based on their specific signal transduction pathway signatures. ERPs unique for ERG (+) and ERG (−) tumors were examined for potential surrogate biomarkers or therapeutic targets based on their cellular localization or enzymatic activity. Consistent with previous reports, we observed the effect of ERG on stimulating cell growth and inhibiting cell differentiation. This is evident in ERG silenced VCaP cells, where we observed increased expression of markers of prostate luminal epithelial differentiation and regulators of cell polarity concomitant with reduced expression of EGFR signaling pathway proteins. Furthermore, to identify correlation to ERG expression at the level of both protein and mRNA expressions, ERPs were compared against the CPDR 80-GeneChip/40-patient tumor vs. benign gene expression dataset (5). Myosin VI and MAOA were found to be positively and inversely correlated to ERG expression, respectively. Combined detection of ERG, Myosin VI and MAOA to distinguish ERG (+) and Myosin VI (+) tumors from ERG (−) and MAOA (+) tumors may facilitate the diagnosis and stratification of CaP patients.

MATERIALS AND METHODS

Cell culture and ERG siRNA knock-down

Human prostate tumor cell lines, VCaP, LNCaP, CWR22Rv1, DU145, PC-3, RWPE-1 and RWPE-2 were purchased from American Type Culture Collection (ATCC, Rockville, MD) and maintained as recommended. The LNCaP subline, C4-2B was purchased from Urocor (Oklahoma, OK) and cultured as recommended. LAPC-4 cells were kindly provided by Dr. Charles L. Sawyers. RC170N cells were established in our laboratory and cultured in Keratinocyte serum-free medium, supplemented with bovine pituitary extract and recombinant epidermal growth factor (Life Technologies, Inc., Carlsbad, CA). (28). ERG (5’-CGACAUCCUUCUCUCACAUAU-3’) and non-targeting (NT; D-001206-13-20) small interference RNA (siRNA) oligo duplexes were from Thermo Scientific (Lafayette, CO) (13). VCaP cells were seeded in 10 cm tissue culture dishes at 2×10⁶ cells per dish in DMEM (ATCC, Rockville, MD), supplemented with 10% charcoal:dextran stripped fetal bovine serum (cFBS; Gemini Bioproducts, West Sacramento, CA) and propagated for three days. Cells were transfected with 25 or 50 nM of ERG or non-targeting (NT) siRNAs using Lipofectamine 2000 (Life Technologies, Inc., Carlsbad, CA) (29). Twelve hours after transfection, VCaP cells were treated with 0.1 nM of the synthetic androgen analogue R1881. A near complete ERG knock-down was achieved by growing the cells for four days following transfection, which was confirmed by immunoblot analysis of the cell lysates.

Prostate tissues and laser capture-microdissection

Under an IRB approved protocol (Protocol No. 20405-28), prostate tumor cells and benign cells (distant to tumor focus) from the same tissue section were isolated by LCM from whole-mounted FFPE prostate sections of five patients that were matched for age (50-65 yrs), race (Caucasian American), and tumor cell differentiation (well to moderate), Gleason grade (3+3 or 3+4) and nuclear grade (grade II). Whole-mounted prostate tissue sections of 8 μm thickness placed on uncharged glass slides were analyzed for malignant and benign cells by hematoxylin and eosin staining and for ERG oncoprotein expression status by IHC with the CPDR anti-ERG monoclonal antibody, ERG-MAb, 9FY (24). Two ERG (+) and three ERG (−) specimens were selected. Approximately 100,000 tumor cells and an equivalent number of matching benign cells were isolated using the Arcturus PixCell II system on LCM caps from each of the sections. The caps were placed into micro-centrifuge tubes with 50 μl of ultra-pure water, immediately frozen on dry ice and stored in −80°C until proteomic analysis.

Proteomic Analysis of ERG Responsive Proteome

ERG (+) tumor cells were pooled together from two specimens; ERG (−) tumor cells, from three specimens; and benign cell, from five specimen. The workflow for proteomic analysis is outlined in Figure 1A. Proteins extracted from the cell pellets were denatured, reduced and alylated before trysinization. Digested peptides were desalted, purified and lyophilized. Peptides were then stacked, resolved and fractionated using capillary isotachophoresis (CITP) and capillary zone electrophoresis (CZE)-based multidimensional separations (26). Peptides fractions were analyzed by nano-reversed-phase liquid chromatography and eluants were monitored by a linear ion-trap mass spectrometer equipped with an electrospray ionization interface. Raw LTQ data were converted to peak list files, which were searched against the UniProt sequence library (www.uniprot.org). A 1% false discovery rate (FDR) for total peptide identifications, which correlates with the maximum sensitivity versus specificity, was chosen as a cutoff. Only proteins identified with at least 2 peptides and 1 unique peptide were included in the final list of identified proteins. The complete description of these procedures is described in detail in the supplementary materials.

Figure 1. Analysis of ERG Responsive Proteome.

(A) Outline of the strategy for analysis of ERG responsive proteome from ERG (+), ERG (−), and benign cells isolated by LCM from prostate cancer specimens and from TMPRSS2-ERG positive VCaP cells. (B) IHC of representative tumor specimens used for LCM with ERG (-) (a & b) and ERG (+) (c & d) expression, stained with H & E and with ERG MAb. (C) Quality control of VCaP cell lysates used in proteomic analysis. Cell lysates were prepared from 50 nM NT siRNA (lane 1) and ERG siRNA (lane 2) transfected VCaP cells and ERG protein was analyzed by immuno-blot analysis with ERG MAb.

Gene Ontology annotation and comparison of datasets

The classification and clustering of proteins dataset were performed using ProteinCenter, v3.2 (Thermo Scientific, West Palm Beach, FL). The differentially expressed proteins detected from the NT siRNA and ERG siRNA experiments were analyzed using the Genomatix (Ann Arbor, MI) GeneRanker and Genomatix Pathway System (GePS) programs. The over-representation of different biological terms (literature association-based or curated canonical pathways) within the input protein list were ranked by their p-values by GeneRanker. Functional interaction networks were generated from these ranked lists based on co-citations within the same sentence in PubMed abstracts linked by a function word. The interaction of ERPs was represented by a network layout that emphasizes high co-citations connectivity and interactions.

Western blot and immunofluorescence assays

VCaP cells were lysed in mammalian protein extraction reagent (M-PER) (Pierce, Rockford, IL) containing protease and phosphatase inhibitors (Sigma, St Louis, MO). Cell lysates equivalent to 20 μg of protein were separated on 4-12% Bis-Tris Gel (Life Technologies, Inc., Carlsbad, CA) and transferred to PVDF membrane. Membranes were incubated overnight at 4°C with primary antibodies and washed before treated with goat anti-Mouse IRDye 800CW or goat anti-Rabbit IRDye 680CW secondary antibodies (Li-Cor Biosciences, Lincoln, NE) at 25°C. Bands were visualized and signal intensities of the bands were quantitatively measured using the Odyssey infra-red imaging scanner (Li-Cor Biosciences).

VCaP cells were seeded onto poly-L-lysine coated coverglass (BD Bioscience; San Jose, CA) in 10% CSS two days prior to siRNA transfection. Cells were induced with 0.1 nM R1881 one day after transfection and incubated for 48 hours. Cells were fixed with PBS buffered 4% paraformaldehyde before permeabilization in 1x PBS with 0.1% Triton X-100. Prior to incubation in primary antibody, cells were blocked in 1% normal horse serum (Vector Laboratories; Burlingame, CA) in PBS. Cells were incubated with a species specific secondary antibody (Alexa-Fluor-594 goat anti-mouse, Alexa-Fluor-488 goat anti-rabbit; Life Technologies, Inc., Carlsbad, CA), and with DAPI (4',6-Diamidino-2-Phenylindole) as a nuclear counterstain .

The antibodies used in for immunoblot and immunofluorescence analysis were acquired from the following sources: ERG-MAb (9FY) from Biocare Medical, Concord, CA ; GAPDH (sc-25778) from Santa Cruz Biotechnology, (Santa Cruz, CA); SHC (610082) from BD BioSciences (San Jose, CA); PAP (2906-1) and ERG (EPR3864(2)) from Epitomics (Burlingame, CA); PSA (A056201-2) and SLC45A3/Prostein (Clone 5E10, M3615) from Dako (Carpinteria, CA); p44/ERK1 (#4372), α-tubulin (11H10, #2125), Cool1/βpix/ARHGEF7 (#4515) and Myosin VI (#9200) from Cell Signaling Technology (Beverly, MA); RAC1 (ARCO3) from Cytoskeleton (Denver, CO); MSMB (TA501072) from Origene (Rockville, MD); Myosin VI (ab126751) and MAOA (EPR7101; ab11096) from Abcam (Cambridge, MA).

Comparison of ERG Responsive Proteome to Gene Expression Datasets

The CPDR 40 patient/80-gene-chip gene expression dataset (GSE32448) was acquired on Affymetrix Human Genome U133 Plus 2.0 arrays using RNA derived from LCM isolated prostate tumors and matching benign tissue specimens. The ERG expression status of the specimens, which were equally represented by moderately-differentiated tumors and poorly-differentiated tumors, were confirmed by IHC and FISH as 14 ERG (−) and 26 ERG (+). Gene expression data of probesets that match ERPs from ERG (+) and ERG (−) tumors were fitted by linear regression model fitting using the lmfit function in the Limma package (30) within the R program. Probesets with most significant correlation or inverse correlation to ERG expression were ranked by eBayes according to the Bayes test statistics in the order differential expression. Statistical significance of a data set was computed using two-tailed t-tests after excluding outliers defined by data-points that are greater than 2.5 standard deviations. Genes and proteins were then compared for concordance in similar up- or down-regulation of tumor vs. normal gene expression ratios to relative MS ratio in ERG (+) and ERG (−) tumors.

RESULTS

Isolation of ERG Responsive Proteins (ERPs)

The strategy to analyze the ERPs in ERG (+) and ERG (−) CaP cells is outlined in Figure 1A. Proteins were isolated from pooled ERG (+) or ERG (−) tumor cells and benign cells from whole-mounted sections of five prostatectomy specimens from patients matched for pathologic stage, age and race (Figure 1B). Proteins were also isolated from NT siRNA and ERG siRNA treated VCaP cells (Figure 1C). Trypsin digested proteins were fractionated by using a CITP/CZE- based multidimensional separations. Peptide fragments were detected with nano-electrospray ionization linear ion trap-tandem mass spectrometry (nano-ESI-MS/MS). The near complete silencing of ERG protein expression was confirmed by immunoblot analysis of ERG siRNA treated VCaP cell lysates (Figure 1C) using ERG-MAb (9FY), which detects ERG protein of 52kDa in VCaP cells (24).

Analysis of the differential expression of ERPs between LCM derived ERG (+) vs. ERG (−) prostate tumor cells

The analysis of ERG responsive proteins isolated from LCM derived ERG (+) and ERG (−) prostate tumor cells and from matched benign cells detected, at 5% false discovery rate (FDR) threshold for total peptide identifications, a combined global proteome of 6171 proteins (Supplementary Figure 1A, Supplementary Table 1A), of which a total of 4,684 were ERPs (Supplementary Figure 1B, Supplementary Table 1B). At stringent threshold for total peptide identifications of 1% FDR, a total of 1,196 ERPs were detected, of which 518 and 500 were unique to ERG (+) and ERG (−) tumor cells, respectively (Figure 2A).

Figure 2.

Global ERG Responsive Proteome detected with at least 2 unique peptide hits and at 1% FDR from LCM isolated tumors (A) and from NT siRNA vs. ERG siRNA transfected VCaP cells (B).

Analysis of the differential expression of ERPs between ERG siRNA vs. NT siRNA transfected VCaP cells

VCaP cells transfected with control NT siRNA oligos showed a robust expression of ERG protein and ERG expression was successfully depleted in the ERG siRNA transfected VCaP cells (Figure 1C). At 5% FDR threshold for total peptide identifications, a total number of 11,416 proteins detected in NT siRNA and ERG siRNA treated VCaP cells (Supplementary Figure 1C, Supplementary Table 1C). At stringent threshold for total peptide identifications of 1% FDR, a combined ERG responsive global proteome of 2,190 proteins was detected. This proteome consisted of 562 proteins detected exclusively in control NT siRNA transfected VCaP cells, 59 proteins exclusively in ERG silenced VCaP cells, and 1,569 differentially expressed proteins common in both NT siRNA and ERG siRNA transfected cells (Figure 2B).

The technical reproducibility of the methods applied for proteomic analyses was verified by performing two independent runs through sequential fractionations by CITP and CZE coupled with LC MS/MS using tryptic digests from VCaP cells that were transfected with NT siRNA and expressing ERG. The reproducibility of the methods used was confirmed by the detection of 80% proteins that were common to two independent runs (Supplementary Figure 1D, Supplementary Table 1 D).

The comparative distribution of the ERG responsive proteins from both NT siRNA and ERG siRNA treated VCaP cells, according to Gene Ontology (GO) instances of defined physiochemical characteristics, including molecular functions, biological processes and cellular compartments, are shown in Supplementary Figure 2. The overall results showed a broad similarity in the range and distribution of proteins from both cells transfected with NT siRNA and with ERG siRNA in the different sub-categories of the GO instances, suggesting robust coverage in the isolation and detection of cellular proteome by the methods employed.

Comparison of ERG Responsive Proteome and ERG Responsive Transcriptome in VCaP cells

We have previously evaluated the transcriptome of VCaP cells in response to ERG knock-down by siRNA using GeneChip microarray analysis (13). Normalized gene expression data from 48 h post transfection were denoted as NT siRNA/ERG siRNA ratios. Comparison of the present set of 2,190 ERPs from VCaP cells (Figure 2B) against probe-sets representing 1,052 distinct genes revealed 250 genes and proteins with concordance response to ERG expression. This represents 23.8% (250/1,052) of the ERG responsive genes from the gene-chip experiments and 11.4% (250/2,190) of the ERG responsive proteins (Supplementary figure 3).

High stringency analysis of ERG Responsive Proteomes of LCM isolated tumor cells and VCaP cells show a strong concordance in regulation by ERG

We compared the proteome of ERPs detected in the LCM isolated prostate tumor specimens (Figure 2A) and in VCaP cells (Figure 2B) to determine the extent of correlation between these two sets of proteome. This evaluation showed an overlap of 489 ERPs, of which 330 ERPs show concordance in their response to up- or down-regulation of ERG protein levels (Figure 3A). The 330 proteins account for 15.1% (330/2,190) of ERPs in VCaP cells and 27.6% (330/1,196) of ERPs in LCM isolated ERG (+) and ERG (−) tumors. The differential levels of detection in ERG (+) vs. ERG (−) tumors and NT siRNA vs. ERG siRNA VCaP cells are shown in Figure 3B.

Figure 3. Overlapping ERG Responsive Proteome of LCM isolated tumors and VCaP cells.

(A) Pie-chart showing 489 ERPs common to ERG (+) vs. ERG (-) tumors and VCaP cells. (B) 330 ERPs concordantly regulated by ERG that are concordantly and differentially regulated by ERG. Dark red and dark green colors represent proteins unique to ERG (+) tumors or NT siRNA treated VCaP cells, and ERG (−) tumors or ERG siRNA transfected VCaP cells, respectively. Lighter shades of red and green represent proteins differentially upregulated or downregulated in these cells. The number of peptides detected for each protein is shown adjacent to each group of ERPs.

Signal transduction pathways of LCM isolated tumor cells and VCaP cells

To evaluate the overall impact of down-stream targets that respond to ERG expression, ERPs from ERG (+) vs. ERG (−) prostate tumor specimens and from VCaP cells were further analyzed using GeneRanker and GePS. ERG responsive proteome networks derived from 1,196 ERPs isolated from tumors and 2,190 ERPs from VCaP cells, as revealed by GePS analysis tool are shown in Figure 4A and 4B, respectively. Proteins that were detected at positive and negative MS ratios for ERG (+) vs. ERG (−) tumor and NT siRNA vs. ERG siRNA in VCaP cells are shown as red and green nodes, respectively. By inference, the red and green nodes represent potential up-and down-regulation by ERG. These networks reflect the impact of ERG expression on protein biosynthesis, chaperone and redox functions, protein trafficking, AR signaling, cell survival and apoptosis, DNA replication, cell cycle control, cell polarity and cell migration. For example, in both ERG (+) vs. ERG (−) tumors and in NT siRNA vs. ERG siRNA treated VCaP cells Proliferating Cell Nuclear Antigen (PCNA) is upregulated, in contrast to prostate specific antigen (PSA /KLK3), which is is downregulated.

Figure 4. Functional interaction networks of ERG Responsive Proteome.

Literature based functional interaction networks of ERPs from LCM isolated prostate tumor cells (A), NT siRNA /ERG siRNA treated VCaP cells (B), and ERPs concordantly regulated by ERG from A and B. Red and green nodes represent proteins unique for ERG (+) and in ERG (−) cells, in the respective samples. Shades of red and green represent upregulated and downregulated ERPs, respectively. In (C), the left- and right-half of the nodes show response to ERG in the LCM isolated prostate tumor cells and in VCaP cells, respectively. Nodes are shown as polygons if the function is known: kinases as right pointed polygons; phosphatases, left pointed polygons; receptors, inverted trapezoids; transporters, trapezoid; and cofactors, stars. Nodes are linked by dotted lines if association is by co-citation and by solid lines if association is by expert curation. (▷) indicates protein A activates protein B; (◇), A modulates B; a circle and bar, A inhibits B; a filled arrowhead, gene B has a binding site for A on one of its promoters.

To highlight the conservation of function in prostate tumors and in the cell culture model, the 330 overlapping ERPs with concordant response to ERG expression in both tumor specimens and cell culture model were analyzed by using GeneRanker and GePS software. The set of 330 overlapping ERPs show, as listed according to p-value rankings in Table 1 and mapped in the resulting network in Figure 4C, an enrichment of pathways regulating cytoskeletal and actin reorganization, as represented by the CDC42-RAC1, the P21 activated protein kinase (PAK) and the actin filaments Y-branching pathways.

Table 1.

Concordantly Regulated ERP Pathways Signal transduction pathways of 330 ERPs concordantly Regulated by ERG. Pathways are ranked by p-values that represent enrichment of proteins of a pathway in the sample.

No	Concordantly Regulated ERP Pathways	Pathway ID	P-value	# Genes (observed)	# Genes (expected)	# Genes (total)	List of observed genes
1	ROLE OF PI3K SUBUNIT P85 IN REGULATION OF ACTIN ORGANIZATION AND CELL MIGRATION	BIOCARTA:CDC42 RAC PATHWAY	6.86E-06	6	0.53	16	ACTR2, ARPC2, ARPC1B, PAK1, ARPC1A, ARPC4
3	P21(CDKN1A) ACTIVATED KINASE	PW_PAK_HOMO_SAPIENS	1.98E-05	9	1.54	68	STMN1, ARPC1B, CALD1, FLNA, PAK2, PAK1, MBP, VIM, PAK3
4	Y BRANCHING OF ACTIN FILAMENTS	BIOCARTA:ACTINY PATHWAY	1.16E-04	5	0.53	16	ACTR2, ARPC2, ARPC1B, ARPC1A, ARPC4
5	PROTEASOME COMPLEX	BIOCARTA:PROTEASOME PATHWAY	1.59E-04	7	1.22	37	PSMB2, PSMC1, PSMD7, PSMA5, PSME2, PSMA6, PSMD3
6	CELL DIVISION CYCLE 42	PW_CDC42_HOMO_SAPIENS	3.28E-04	9	2.20	97	DNM2, GNA13, ARHGAP1, INF2, RALA, PAK2, PAK1, VIM, PAK3
7	ER ASSOCIATED DEGRADATION (ERAD)	BIOCARTA:ERAD PATHWAY	2.95E-03	4	1.22	19	MAN2B1, MOGS, CANX, GANAB

ERG knock-down induces the expression of prostate differentiation markers associated with its secretory function and impacts the epidermal growth factor receptor (EGFR) signaling pathway

In our earlier publication, we have noted that ERG interferes with prostate epithelial differentiation by inhibiting a number of genes including KLK3, SLC45A3 (Prostein), C15ORF21 (Dresden prostate carcinoma 2 protein (D-PCa-2)) and MSMB (β-microseminoprotein/PSP94) (13). In the current study, we detected a consistent expression pattern of these proteins in relation to ERG expression in both the LCM isolated tumor cells and VCaP cells with the ERG knock-down. The protein expression and sub-cellular localization of several ERG responsive downstream targets, were validated in VCaP cells following ERG siRNA treatment. In response to ERG knock-down the expression of cytoplasmic SLC45A3 and prostatic acid phosphatase (PAP/ACPP) were dramatically upregulated (Figure 5A and B), but MSMB expression showed a more subtle increase (Figure 5C), consistent with the results of ERG responsive transcriptome. The upregulated expressions of Prostein, PSA, and PAP/ACPP in response to the ERG siRNA in VCaP cells were also validated by immunoblot analysis (Figure 5D).

Figure 5. ERG knock-down induces the expression of prostate differentiation markers associated with its secretory function.

Validation of the upregulated expression prostate differentiation markers, (A) SLC45A3, (B) PAP/ACPP and (C) MSMB in VCaP cells upon ERG siRNA by immunofluorescence assay and by immunoblot analyses (D).

Markers of cell growth and proliferations from the epidermal growth factor receptor (EGFR) signaling pathway, such as the Src homology 2 domain containing transforming protein 1 (SHC1) and mitogen-activated protein kinase (p44/ERK1) (31), show higher levels of expression when ERG is expressed in the cell, but becomes down-regulated when ERG expression is silenced by siRNA (Figure 6A-B). In contrast, the expression of regulators of cell polarity and apical junction assembly, such as Rho-GTPase, RAC1 (32) and Rho guanine nucleotide exchange factor 7 (ARHGEF7/p85 Cool1/βPix) (33) is elevated in response to ERG knock-down (Figure 6C-D), which confirms the inhibition of prostate epithelial differentiation by ERG. The downregulation of SHC1 and p44/ERK1 and upregulation of ARHGEF7 were also validated by immunoblot assays (Figure 6E).

Figure 6. ERG knock-down inhibits genes regulating cell growth and activates genes regulating prostate epithelial differentiation.

EGFR pathway proteins (A) SHC1, (B) p44/ERK1 are down-regulated by ERG siRNA. The silencing of ERG increased the expression of the Rho-GTPase RAC1 (C), and ARHGEF7 (D). The down-regulation of SHC1 and p44/ERK1 and up-regulation of ARHGEF7 are confirmed by immunoblot analysis.

Signal transduction pathways signatures defined by ERG (+) and ERG (−) tumors

The identification of ERPs that are exclusive to or overexpressed in either ERG (+) or ERG (−) tumors could further reveal functional roles of ERG in prostate tumor initiation and progression. Proteins that are correlated with ERG expression could serve as surrogate biomarkers and/or therapeutic targets in ERG (+) tumors. In contrast, proteins that are overexpressed in tumors lacking ERG could be used as biomarkers that define a separate category of tumors. The relative abundance of a protein in ERG (+) vs. ERG (−) tumors, or in ERG (+) and ERG (−) tumors vs. benign tissues was determined based on its relative MS ratio. Since the same protein could be detected in these separate analyses, we sorted the proteome data again to identify proteins that were detected at higher ratios in ERG (+) or in ERG (−) tumors, relative to benign tissues. 589 proteins were detected at higher ratios in ERG (+) tumors, of which 204 were unique for ERG (+) tumors. Conversely, 504 of the 781 proteins detected at higher ratios in ERG (−) tumors, were exclusively for ERG (−) tumors (Figure 7).

Figure 7. Identification of ERPs detected at higher ratios in ERG (+) or in ERG (−) cells.

204 of 589 ERPs detected more abundantly in ERG (+) relative to ERG (−) tumors or benign tissues are unique to ERG (+) tumors. 504 of 781 ERPs detected more abundantly in ERG (−) relative to ERG (+) tumors or benign tissues are unique to ERG (−) tumors.

To further identify the individual profiles that define the proteome from ERG (+) and ERG (−) prostate tumors, ERPs from each set were analyzed for pathway enrichment and associated literature-based networks using GeneRanker and GePS. Analysis of ERPs from ERG (+) tumors revealed, as listed according to p-value rankings in Table 2A, enrichment for pathways that regulate cell shape and motility (PAK pathway), remodel cytoskeletal structure (CDC42 pathway), promote cell survival (AKT pathway), and enhance protein synthesis and cell growth (AKT-MTOR pathway). The nodes that connect these pathways include MTOR and GSK3B (Figure 8A). ERPs from ERG (+) tumors which are localized to the plasma membrane or released into the extracellular compartments include: P21 protein activated kinase 1 (PAK1); synaptotagmin1 (SYN1), a regulator of exocytosis; components of the clathrin-mediated endocytosis, epidermal growth factor receptor pathway substrate 15 (EPS15) and dynamin 1 (DMN1); S100 calcium binding protein A13 (S100A13) and glutathione peroxidase 3 (GPX3) (Figure 8A).

Table 2A.

Signal Transduction Pathways for 204 ERPs unique for ERG (+) tumors.

No	ERG (+) ERP Pathways	Pathway ID	P-value	# Genes (observed)	# Genes (expected)	# Genes (total)	List of observed genes
1	P21 (CDKN1A) ACTIVATED KINASE	PW_PAK_HOMO_SAPIENS	1.55E-05	8	1.14	68	PAK2, SYN1, PAK1, WASF2, STMN1, PAK4, CALD1, PAK3
2	CELL DIVISION CYCLE 42	NCI-Nature:CDC42	9.36E-04	6	1.17	74	GSK3B, PAK2, PAK1, MTOR, EPS8, PAK4
3	V AKT MURINE THYMOMA VIRAL ONCOGENE HOMOLOG 1	PW_AKT1_HOMO_SAPIENS	2.03E-03	11	3.99	238	PPM1G, GSK3B, PDCD4, DNAJC5, PAK1, HTT, PITPNA, PHLDB1, MTOR, UTRN, SPARC
4	AKT-MTOR	BioCarta:IGF1-MTOR	3.85E-03	6	1.51	90	SYN1, HTT, ITPR1, STMN1, EPS8, RAB11FIP5

Figure 8. Functional interaction networks of ERPs unique for ERG (+) and ERG (-) tumors.

Red and green nodes represent proteins upregulated in ERG (+) tumors (A) and ERG (-) tumors (B), in the respective samples.

Pathways that were found to be enriched for ERPs derived from ERG (−) tumors, as listed according to p-value rankings in Table 2B, function in the proteolytic degradation of proteins (proteasome pathway), integrin-mediated cell migration (mammalian calpain pathway), MAP kinase pathway activation via G protein coupled receptors (GPCR pathway), actin remodeling and cell migration (CDC42-RAC pathway), and the prevention of oxidative damage of proteins (redox pathway). The nodes that connect these pathways function in the control of cell motility (CDC42 and Calpain II (CAPN2)), proteolysis of extracellular matrix (plasminogen (PLG)), fatty acid metabolism (adiponectin (ADIPOQ)), and signal transduction at the caveolae scaffolding of plasma membrane (caveolin I (Cav1)) (Figure 8B).

Table 2B.

Signal transduction pathways 504 ERPs unique for ERG (−) tumors.

No	ERG (−) ERP Pathways	Pathway ID	P-value	# Genes (observed)	# Genes (expected)	# Genes (total)	List of observed genes
1	PROTEASOME COMPLEX	BIOCARTA:PROTEASOME PATHWAY	5.49E-07	11	1.75	37	PSMA3, PSME1, PSMA4, PSMB2, PSMC1, PSMD7, PSMA5, PSMC4, PSME2, PSMD6, PSMB4
2	MCALPAIN AND FRIENDS IN CELL MOTILITY	BIOCARTA:MCALPAIN PATHWAY	7.73E-06	9	1.46	31	PRKAR2B, MAP2K2, CAST, PRKAR1A, CAPN2, ITGA1, PRKAR2A, MAP2K1, MAPK3
3	SIGNALING PATHWAY FROM G-PROTEIN FAMILIES	BIOCARTA:GPCR PATHWAY	2.15E-05	8	1.28	27	PRKAR2B, GNAI1, PRKAR1A, PPP3CA, ASAH1, PRKAR2A, MAP2K1, MAPK3
4	ROLE OF PI3K SUBUNIT P85 IN REGULATION OF ACTIN ORGANIZATION AND CELL MIGRATION	BIOCARTA:CDC42 RAC PATHWAY	5.47E-05	6	0.76	16	ACTR2, ARPC2, ARPC1B, CDC42, ARPC1A, ARPC4
5	REDOX	PW_REDOX_HOMO_SAPIENS	8.55E-05	14	4.28	126	HSPA4, H6PD, CSNK2A1, AKR1B1, MPST, AKR1A1, NAMPT, TXNDC17, ADIPOQ, PSME2, SOD2, TXNRD1, PRDX4, GPX1

Association of Myosin VI (MYO6) and Monoamine oxidase A (MAOA) to ERG mRNA and protein expression

In order to identify biomarkers that are tightly regulated by ERG, both at the level of protein expression and gene expression, we compared ERPs that are detected at higher ratios in ERG (+) or in ERG (−) tumors against the CPDR tumor vs. normal 80-GeneChip gene expression dataset (GSE32448) of tumor vs. normal gene expression ratios from 14 ERG (−) and 26 ERG (+) cases (5). ERPs were matched against the probesets in this dataset and linear regression model fitting was used to identify genes that most closely correlate to or inversely correlate to ERG expression. By ranking the probesets according to their differential expression, MYO6 was identified to correlate most closely to the gene expression profile of ERG across 40 patients (p-value = 3.92E-06) (Table 3A, Figure 9A, B and D). The probability of differential expression of MYO6 in ERG (+) and ERG (−) tumors relative to ERG was 0.92. There is a 3.6-fold difference between the means of gene expression ratios between ERG (+) vs. ERG (−) cases for MYO6. On the contrary, the expression of MAOA is noted have the most statistically significant inverse correlation to ERG expression (p-value=0.0134) and a high probability to be mutually exclusive to ERG (0.90). (Table 3B, Figure 9C and D).

Table 3A.

Probe sets for MYO6 revealed gene expression profiles that are most significantly correlated to ERG Expression.

Symbol	Probesets	Rank coeff=2, ERG (+), Lg2 scale	B-Statistic	Probability gene is Differentially Expressed	T-test	Med ERG (+) -Med ERG (−)	Ave ERG (+) -Ave ERG (−)
ERG	213541_s_at	1	12.19	1	3.21E-11	4.11	3.51
MYO6	203215_s_at	3	2.49	0.92	3.92E-06	1.76	1.66
CSDA	201160_s_at	21	−2.78	0.06	2.16E-03	0.99	1.29
UTRN	225093_at	25	−3.02	0.05	7.45E-03	0.81	1.03

Figure 9. Correlation of gene and protein expression between ERG with Myosin VI and MAOA.

Box-plots showing the range of log 2 tumor vs.normal expression ratio from 40-patient gene expression dataset for ERG (213541_s_at) (A), MYO6 (203215_S_at) (B) and MAOA_204389_at (C) according to ERG expression status. The line across the box and the blue spot represent the respective median and mean values, respectively. The relative expression of ERG, MYO6 and MAOA are normalized by row or Z-score (D, top) or shown as original values (D, bottom). The correlation of Myosin VI and MAOA expression to ERG expression is validated in ERG siRNA knockdown of VCaP cells and assayed by immunoblot analysis (E) and immunofluorescence assay (F). Expression of Myosin VI and MAOA were compared in CaP cell lines (G). Induction of VCaP and LNCaP cells with 0.1 nM and 1 nM of R1881 for 12, 24 and 48h following growth in starvation conditions for three days.

Table 3B.

Probe sets for MAOA revealed gene expression profiles that are most significantly inversely correlated to ERG Expression.

Symbol	Probesets	Rank coeff=1 (ERG−) Lg2 scale	B-Statistic	Probability gene is Differentially Expressed	T-test	Med ERG (+) -Med ERG (−)	Ave ERG (+) -Ave ERG (−)
MAOA	204389_at	3	2.16	0.9	1.34E-02	−0.446	−0.372
IMMT	242361_at	45	−2.92	0.05	1.04E-02	−0.344	−0.467
NEDD4L	241396_at	83	−3.65	0.03	1.41E-02	−0.338	−0.321
ERG	213541_s_at	2049	−5.93	0	3.21E-11	4.109	3.510

Verification of Proteomic and Gene expression Data for Myosin VI and MAOA

The correlation of Myosin VI and inverse correlation of MAOA to ERG expression were further validated in independent ERG siRNA experiments using VCaP cells. The silencing of ERG in comparison to the control experiment resulted in a reduction of Myosin VI expression by half compared to a two-fold increase of MAOA expression (Figure 9E), as measured by quantitative evaluation of the immunoblot intensities. Immunofluorescence assay confirmed the diminished expression of Myosin VI expression and upregulation of MAOA in response to ERG siRNA (Figure 9F). The response of MAOA expression to ERG protein levels corroborates with the data from mass spectrometric analysis, which show MAOA to be detected exclusively in ERG (−) tumors and at approximately 3-fold higher MS ratios in ERG siRNA vs. NT siRNA treated VCaP cells (Figure 3B).

In order to find out the status of Myosin VI and MAOA expression in other CaP cells that do not express ERG, we examined the expression of these two proteins in a panel of CaP cell lines (Figure 9G). In addition to VCaP cells, Myosin VI was shown to be strongly expressed in LNCaP cells, which lacks ERG expression. Interestingly, the expression of MAOA was shown to be associated with the level of AR expression in AR positive LNCaP, C4-2B and CWR22rv1 cells. We next examined the regulation of Myosin VI and MAOA by androgen using LNCaP cell and VCaP cells grown under starvation conditions and induced with 0.1 nM and 1 nM of R1881 (Figure 9H). Myosin VI expression level did not show any detectable response to R1881 in either LNCaP or VCaP cells. However, a two-fold increase for MAOA expression in LNCaP cells was observed after 48 h induction, both at 0.1 nM and at 1 nM, as measured from the immunoblot intensities (Figure 9H, right panel). In contrast, MAOA expression is downregulated concomitant with the R1881 induced ERG expression in VCaP cells (Figure 9H, left panel), which confirms the inverse correlation gene expression between MAOA and ERG.

Meta-analysis of the association of gene expression between ERG, MYO6 and MAOA in independent datasets

The association of gene expression between ERG with MYO6 and MAOA was further evaluated in two independent CaP gene expression data sets with known TMPRSS2-ERG gene fusion status: the study involving the Swedish watchful waiting cohort (GSE16560) (34) and the Memorial Sloan-Kettering Cancer Center (MSKCC) study using primary and metastatic tumors (GSE21032) (35). ERG fusion status was available in 272 of the 281 cases from the Swedish watchful waiting cohort by FISH analysis. In the MSKCC cohort, ERG fusion status was confirmed for 128 cases by array comparative genomic hybridization (aCGH) data. Patient data were clustered according to ERG fusion status and the ranges for log 2 gene expression for ERG fusion positive and ERG fusion negative were plotted as box and whisker plots. MYO6 expression was significantly higher in ERG (+) tumors versus ERG (−) tumors in both the Swedish watchful waiting cohort (p-value = 1.71 E-6) and the MSKCC cohort (p-value = 1.40 E-7), confirming that MYO6 expression is significantly correlated with TMPRSS2-ERG fusion or ERG over-expression. However, a similar evaluation of MAOA expression did not show any inverse correlation as observed in this study.

Discussion

The discovery of ERG overexpression in prostate tumors and the fusion of genes involving TMPRSS2 promoter region with the ERG coding sequences in more than 50% of CaP has opened avenues for exploration of biomarkers useful for the detection and the stratification of CaP. The PSA assay used in the clinical screening of CPa is known to lack both sensitivity and specificity (36). Therefore there is a need to identify biomarkers that have the potential to not only accurately detect clinically relevant CaP in asymptomatic patients but also able to differentiate indolent from aggressive CaP. Analysis of ERG (+) and ERG (−) tumor specimens is likely to provide additional information about novel biomarkers for potential clinical use.

We analyzed ERG responsive proteome in CaP cells with the aim to understand the function of ERG in the etiology of CaP and to identify biomarkers that are associated with ERG (+) or ERG (−) status of the prostate tumor. A notable feature of this study is the typing of tumors for ERG expression by IHC followed by targeted selection by LCM to overcome the challenges imposed by the presence of ERG (+) and ERG (−) tumor foci in the same prostate. Hence, a straightforward comparative analysis involving ERG (+) and ERG (−) tumors and normal cells is likely to show the potential differences at the proteome level. Unlike other CaP-proteomics studies (18,20,37,38), the current study focused specifically on the detection of proteins that are differentially expressed in relation to ERG oncoprotein status. To address this, we have utilized a unique proteomics platform based on CITP/CZE multidimensional separation coupled with nano-ESI-MS/MS, which involve minimal front end purification prior to mass spectrometry analysis. The ability to reproducibly detect over 80% of the same peptides in consecutive runs using aliquots of the same protein samples demonstrated the reliability of the techniques used (Supplementary Figure 1D).

In addition to normal prostate epithelial cells and ERG stratified tumor cells, we have also analyzed the ERG responsive proteome of VCaP cell line. Such an analysis revealed a distinct pattern of up and downregulation of proteins in response to ERG that was corroborated by concordance to mRNA expression reported by previous gene expression analyses (13). The detection of similar responses in protein and mRNA expression in protein markers of prostate luminal epithelial differentiation and secretory function to ERG siRNA knock-down further confirmed the reliability of the methods used in this proteomic analysis. Examples of these proteins, which are shown in Figure 3 and in Supplementary Figure 3, include KLK3 (PSA), SLC45A3, TMPRSS2 and prostate-specific membrane antigen-like protein (FOLH1B), semenogelin-2 (SEMG2) and transgelin (TAGLN).

ERP highlights overlapping pathways in ERG (+) and ERG (−) prostate tumors and in VCaP cell line

The characterization of ERG function through analysis of interaction networks based on ERG responsive proteome datasets captured a representation of previously reported ERG target genes (11-13,16). The overlapping signal transduction networks revealed for ERG (+) and ERG (−) prostate tumors and NT siRNA and ERG siRNA treated VCaP cells are consistent with the activation of cell growth and cell proliferation and the inhibition of prostate epithelial differentiation by ERG oncoprotein (Figure 4A and B).

A more concise signal transduction network of literature-based interactions of the ERP was generated from the 330 proteins concordantly regulated by ERG in both tumors and in VCaP cells. The central nodes of this network (Figure 4C), represented by vimentin (VIM) and albumin (ALB) highlights the role ERG plays in regulating modulators of glandular prostate epithelial differentiation and secretory function. Vimentin is an intermediate filament protein that is expressed early during cell differentiation, promotes cell invasiveness, is expressed by motile prostate cell lines and positively correlates with poorly differentiated cancers and bone metastases (39). Albumin acts as a carrier protein for steroids, fatty acids, and thyroid hormones and functions to stabilize extracellular fluid volume in body fluids including prostatic fluids (40). This network complements the signal transduction pathways from GeneRanker analysis, which show an enrichment of pathways regulating cytoskeletal and actin reorganization, cell migration, protein biosynthesis and proteasome and ER-associated protein degradation pathways (Table 1). These pathways underscore the association between structure and function during prostate epithelial differentiation or epithelial-mesenchymal-transition (EMT). During these events, changes in cell shape and polarity occur simultaneously with changes in the expression of protein markers of prostate epithelial differentiation or EMT. For example, the dynamic response to ERG expression by Rho-GTPase CDC42-RAC1 signaling related pathways that regulate actin filament dynamics (Figure 6C and D) is accompanied by equally robust alterations to AR function evident from in the dysregulation of PSA, SLC45A3, and PAP/ACPP (Figure 5).

ERPs from ERG (+) and ERG (−) tumors reveal distinct protein markers and signature pathways

The disparity in ERG fusion frequency among different ethnic populations points to yet undiscovered genetic alterations that may contribute to the initiation and progression of CaP. We have defined the ERP by ERG expression status to better understand the biological features that distinguish these two classes of tumors. The different profiles of the ERP from ERG (+) and ERG (−) tumors, while highlighting the role ERG plays in regulating diverse cellular functions, may reveal distinctive signatures that could help to stratify ERG (+) from ERG (−) tumors and discover new treatment options. The identification of 204 and 504 ERPs unique for ERG (+) and ERG (−) tumors, respectively, represent a distinct and informative subset of the ERPs (Figure 7). The connection of PAK, CDC42 and AKT pathways enriched in ERG (+) tumor derived ERPs by MTOR and GSK3B, underscore the impact that ERG overexpression may have on the PI3K/AKT/mTOR, the PI3K/AKT/GSK3B or Wnt signaling pathways (Table 2A, Figure 8A). The pathways enriched in ERPs from ERG (−) tumors regulate functions that include CDC42-RAC and calpain modulated cell motility and proteasome and redox functions. These pathways were shown to be connected by CDC42, CAV1, SOD2, ADIPOQ and MAPK3 (Table 2B, Figure 8B). These links reveal that in addition to changes in cell differentiation and migration, cell survival and apoptosis, the absence of ERG in these tumors also affect changes in endocytosis and protein trafficking , redox and proteasome functions, as well as fatty acid metabolism.

The PI3K/AKT/mTOR and the MAP kinase pathways have been implicated in CaP tumorigenesis and development of castrate resistant prostate cancer, and targeting these pathways to treat CaP using small molecule inhibitors is an active area of investigation (41,42). Several of the ERG (+) and ERG (−) specific ERPs, which are identified to be secreted in body fluids or found localized to the plasma membrane warrant more detailed studies to evaluate their potential as diagnostic protein biomarkers or as targets for treatment for CaP.

Correlation of Myosin VI and Monoamine oxidase A with ERG gene and protein expression

The combined analysis of proteomic and genomic data for proteins positively correlated and inversely correlated to ERG expression identified Myosin VI and MAOA as potential protein and gene expression biomarkers for ERG (+) and ERG (−) tumors. Myosin VI is one of the unconventional myosins, actin-based molecular motors involved in intracellular vesicle and organelle transport. Although Myosin VI has previously been reported to be over-expressed in CaP (43), this is the first report of a correlation with ERG expression in CaP. The localization of Myosin VI on endosomes and the trans-Golgi network, suggest a function in regulating protein secretion (44,45). Myosin VI has been implicated in autophagy by promoting autophagosome maturation and driving fusion with lysosomes (46). The correlation of Myosin VI gene and protein to the expression of ERG suggests possible transcriptional modulation of Myosin VI by ERG.

MAOA is a mitochondrial enzyme expressed in the brain and peripheral tissues that degrades biogenic amines including neurotransmitters serotonin and norepinephrine by oxidative deamination, resulting in the production of hydrogen peroxide (47). In normal prostate glands MAOA is absent or found at very low levels in the luminal secretory epithelial but is elevated in the basal epithelia (48). Increased MAOA expression is also found to be associated with poorly differentiated high grade CaP (49) while a rare polymorphism of the MAOA promoter that confers low expression was associated with reduced CaP risk (50). In this study, both gene and protein expression of MAOA are found to be expressed at higher ratios in ERG (−) tumors compared to ERG (+) tumors. MAOA was detected at almost 3-fold higher MS ratios in ERG silenced VCaP cells compared to NT siRNA control (Figure 3B). In the context of prostate epithelium, MAOA is regulated by androgens through promoter-upstream glucocorticoid/androgen response elements (51). We showed that although MAOA is upregulated by R1881 induction in ERG (−) LNCaP cells, the expression of ERG in VCaP cells appear to interfere with this activation. The inverse correlation of MAOA with ERG suggests that MAOA may define a separate and distinct category of ERG (−) but androgen sensitive tumors. Although the primary function of MAOA is the oxidative deamination of monoamine neurotransmitters, whether the overexpression of MAOA in prostate epithelium leads to the oxidative deamination of prostatic polyamines such as spermine or spermidine, and the release of reactive oxygen species that contribute to tumorigenesis, remains to be shown.

A comparison of ERG, MYO6 and MAOA expression to independent gene expression datasets from Sboner et. al., (34) and Taylor et. al., (35) confirmed the significant correlation of MYO6 but not the inverse correlation of MAOA with ERG. This could be attributed the differences in procedures used for mRNA sampling from tissues or in the sensitivity of microarray platforms. Unlike the two larger datasets, which used only tumor mRNA, the 80-GeneChip dataset analyzed in this study used mRNA from both tumor and normal cells. The preparation of the mRNA from LCM isolated cells further reduces heterogeneity or contamination of non-tumor cells.

Potential treatment of CaP based on the stratification for ERG, Myosin VI and MAOA expression

The prevalence TMPRSS2-ERG fusion and its function as a driver mutation in the initiation and progression of CaP present a promising therapeutic target. Transcription factors such as ERG were considered “undruggable”, mainly due to its inaccessibility. Nevertheless inhibition of ERG function exemplified by the use of small molecule inhibitors and TMPRSS2-ERG fusion junction specific siRNAs have been successfully carried out with varying degree of success (52). The inhibition of the DNA dependent interaction of ERG with poly(ADP-ribose) polymerase (PARP) with PARP inhibitors (53) has advanced rapidly due to the availability of pharmacological inhibitors. The development of combination assays using triple immunostaining cocktails and /or nucleic acid detection panels could help categorize tumors according to their expression of ERG, Myosin VI or MAOA. Parallel advances in the development of specific small molecule inhibitors, used either alone or in a combinatorial approach with other drugs, could be applied to synergistically inhibit ERG (+) or ERG (−) tumors. The small molecule inhibitor, 2,4,6-triiodophenol has been recently shown to reduce the number of Myosin VI-dependent vesicle fusion events at the plasma membrane during constitutive secretion (54) and could be used to inhibit the formation of autophagosomes (46). Since the genetic alterations that define ERG (−) tumors are not well understood, the identification of ERPs overexpressed in ERG (−) tumors or are expressed in inverse correlation to ERG, such as MAOA, could help to uncover the mechanisms responsible for the initiation and progression of ERG (−) CaP. The therapeutic potential of MAOA inhibitors such as clorgyline, which induces differentiation in primary cultures of normal basal epithelial cells and high-grade CaP, is being actively investigated (55). Further developments in this direction could translate to the development of clinical treatments for CaP patients based on the ERG expression status of the tumors.