ERG oncoprotein inhibits ANXA2 expression and function in prostate cancer

Abstract

Overexpression of ERG in the prostate epithelium, due to chromosomal translocations, contributes to prostate tumorigenesis. Here, genomic analysis of ERG siRNA-treated prostate cells harboring the endogenous TMPRSS2-ERG fusion revealed an inverse relationship between ERG and Annexin A2 (ANXA2) expression at both the RNA and protein level. ANXA2, a Ca(2+)-dependent and phospholipid-binding protein, is involved in various cellular functions, including maintenance of epithelial cell polarity. Mechanistic studies defined the prostate-specific transcription start site of ANXA2 and showed that the recruitment of ERG to the ANXA2 promoter is required for transcriptional repression by ERG. Knockdown of ERG enhanced the apical localization of ANXA2, the bundling of actin filaments at cell-cell junctions and formation of a polarized epithelial phenotype. ERG overexpression disrupted ANXA2-mediated cell polarity and promoted epithelial-mesenchymal transition (EMT) by inhibiting CDC42 and RHOA, and by activating cofilin. Immunohistochemistry demonstrated a reciprocal relationship of ANXA2 and ERG expression in a large fraction of primary prostate cancer clinical specimens. ANXA2 was absent or markedly reduced in ERG(+) tumors, which were mostly well differentiated. ERG(-) tumors, meanwhile, expressed moderate to high levels of ANXA2, and were either poorly differentiated or displayed subsets of poorly differentiated cells. Taken together, the transcriptional repression of ANXA2 by ERG in prostate epithelial cells plays a critical role in abrogating differentiation, promoting EMT, and in the reciprocal correlation of ERG and ANXA2 expression observed in human prostate cancer.

Implications: ANXA2 is a new component of the ERG network with potential to enhance biologic stratification and therapeutic targeting of ERG-stratified prostate cancers.

Figure 1. Gene expression analysis of 40 patients (20 WD and 20 PD) showed inverse correlation of gene expression between ERG and ANXA2

Mean differences of mRNA expression between WD and PD, ERG(+) and ERG(−), as well as normal and tumor samples for members of the ANNEXIN family (A). Tumor vs. normal expression ratios ofERG and ANXA2 in 20 WD and 20 PD tumors showed an inverse correlation (B). Results from GeneChip analysis of VCaP cells, 24 and 48 h after ERG siRNA knock-down showing ratios of ANXA2 expression for ERG/NT siRNA, averaged from three ANXA2 probe sets, 213503_x_at, 201590_x_at and 210427_x_at. (Two-tailed Student’s T-test, *p- value=6.38×10⁻⁶; **p-value=4.51×10⁻⁵) (C). Immunoblot assay of lysates from VCaP cells 72 h after transfection with 25 or 50 nM of ERG, ANXA2 or combined ERG and ANXA2 siRNA oligos (D).

Figure 2

Mapping of prostate specific ANXA2 transcript, analysis of ERG binding and transcription regulation ANXA2 promoter. 5’RACE analysis identified a prostate specific transcript(A). The 5’ end of the transcript is spliced together from exons 1 and exon 2 (red triangles) of ANXA2 mRNA, variant 1 (NM_001002858.2) and variant 3 (NM_004039.2), and is translated starting from exon 2 (B). ChIP of chromatin from NT and ERG siRNA treated VCaP cells using ERG MAb showed relatively strong recruitment of ERG to V$ETSF motifs #1, #2, #3, #6 and #7. The ratio of immunoprecipitated chromatin to input chromatin is shown adjacent to each amplified region. The recruitment of ERG to the C-MYC and HPGD promoter regions were evaluated as positive controls and a region upstream of ANXA2 promoter was tested as a negative control (C). Analysis of luciferase reporter constructs containingANXA2 promoter upstream sequences transfected indoxycycline-inducible ERG expressing LNCaP-LTE3 cell line (D, upper; *, p=value < 0.05, **, p=value < 0.05) and in NT and ERG siRNA treated VCaP cells (D, lower; †, p=value < 0.05). Mean relative luciferase activities were calculated from triplicate transfections.

Figure 3. ERG knock-down induces ANXA2 overexpression

VCaP cells were transfected with 50 nM of ERG, ANXA2 or combined ERG and ANXA2 siRNA oligos and analyzed after 72 h by immunofluorescence assay (A). XZ- and YZ-sections, taken through the region depicted by the green and red lines, respectively, show that ERG siRNA knock-down induces accumulation of ANXA2 at the apical surface of VCaP cells (white arrowhead) compared to control NT siRNA (B). Images of TIRF field showing control and ERG siRNA knock-down cells. Signal intensity is averaged from 4 sets of 15 cells each (unpaired t- test; n = 4 sets of 15 cells each; p-value = 0.032) (C). All bars represent 10 µm.

Figure 4

A. The knock-down of ERG promotes the reversal of EMT and reestablishes the polarized epithelial phenotype of the VCaP cells, demonstrated by the over-expression of determinants of the polarized epithelial cells and components of the AJ (E-cadherin and beta-catenin) and Tight Junction (ZO-1). B. Compared to control NT-siRNA, knock-down of ERG increases organized f-actin polymerization, while knock-down of ANXA2 results in disorganized f-actin network. Cells were fixed and stained with Alexa-594 phalloidin 72 h after transfection. C. ERG inhibits the CDC42 and RHOA activity through repression of ANXA2. The relative activity of activated GTP bound forms of (i) CDC42 (ii) RHOA and (iii) RAC1 in were measured by G-LISA assays 72 h after siRNA transfection (C)((Two-tailed Student’s T-test: *, P < 0.05; **, P < 0.05; ***, P < 0.05; ^†, P < 0.05;^††, P < 0.05;^†††, P < 0.05; †, P < 0.05; ††, P < 0.05). The expression of phosphorylated cofilin is enhanced by ERG knock-down (lanes 3 and 4), but is reduced by ANXA2 (lanes 5 and 6) and by combined ERG and ANXA2 knock down (lanes 7 and 8) (D). ERG regulates the transition between the polar differentiated epithelial phenotype and the mesenchymal phenotype by inhibiting CDC42 and RHOA mediated actin-polymerization through the repression of ANXA2 (E).

Figure 5

IHC staining of consecutive sections of representative whole-mounted prostate sections with H & E, ERG and ANXA2 antibodies demonstrate the inverse correlation of ERG and ANXA2 in prostate adenocarcinoma. (A) A representative whole-mounted prostate section with index tumor of ERG(+)/ ANXA2(−) phenotype. In the second row, the region indicated by the rectangle is magnified, showing membranous and cytoplasmic staining of ANXA2 in the luminal and basal cells of benign prostate epithelium (blue arrows). Black arrows indicate well differentiated (WD) ERG(+)/ANXA2(−) cells. (B) WD tumors with both ERG(+)/ANXA2(−) (black arrows) and ERG(−)/ANXA2(−) (red arrow) phenotype. Blue arrow indicates benign gland. (C) and (D) represent poorly differentiated (PD) tumors with inversely correlated ERG(+)/ANXA2(−) and ERG(−)/ANXA2(+) phenotype, respectively. (E) Prostatic ductal adenocarcinoma (papillary variant) which is ERG(−) but has both ANXA2(−) and ANXA2(+) cells. WD and PD tumors of ERG(−)/ANXA2(−) phenotype are represented in (F) and (G), respectively. Benign glands (blue arrows) and endothelial cells (green arrowheads) are ANXA2(+).