Activation of NF-κB by TMPRSS2/ERG Fusion Isoforms through Toll-Like Receptor-4 Free

The discovery of recurrent fusion of the androgen-regulated TMPRSS2 gene to the ETS transcription factors (1), particularly the ERG gene, in the majority of prostate cancer (PCa) lesions, has led to intensive study of the biological and clinical consequences of these rearrangements in PCa. The TMPRSS2/ERG (T/E) fusion gene is by far the most common fusion gene and is present in 40% to 60% of PCas (2, 3). In all cases, the TMPRSS2 promoter and one or more TMPRSS2 5′ exons are juxtaposed to the ERG exons, with deletion of the ERG promoter. Therefore, the fusion transcripts are under the control of the androgen-regulated TMPRSS2 promoter, resulting in the constitutive high level expression of ERG fusion transcripts in the neoplastic prostatic epithelium bearing this fusion gene. ERG shares a conserved 85 amino acid DNA-binding domain with other ETS factors, which are generally oncogenic and promote tumor progression in a number of malignancies.

There is significant heterogeneity in the structures of the T/E fusion transcripts, both at the 5′ end of the mRNA (2, 3) and downstream ERG coding exons (4). We previously characterized 8 fusion isoforms based on the fusion junction structure (3). The most common transcript contains the TMPRSS2 exon 1 fused to ERG exon 4, which we have designated as the type III isoform (3). This variant is expressed in the vast majority (80%–90%) of fusion gene expressing PCa, either alone or in combination with other isoforms. Translation would have to arise from an internal ATG codon because the native ERG ATG is absent and would thus give rise to a slightly truncated ERG protein. Of particular interest is an isoform in which TMPRSS2 exon 2 is fused with ERG exon 4 (designated type VI). This variant was present in 26% of the cases with T/E fusion gene expression (3). For this isoform, the in-frame fusion results in a true fusion protein containing the first 5 amino acids of the TMPRSS2 gene fused to a slightly truncated ERG protein. We found that expression of this isoform is associated with aggressive disease. Subsequent studies have shown that the type VI isoform enhances proliferation and invasion of immortalized prostatic epithelial cells to a greater extent than the type III isoform, consistent with the clinical associations (4). In addition, we have identified multiple alternative splicing exons of ERG by cloning and sequencing of the fusion gene transcripts (4). In particular, alternative splicing leading to inclusion or exclusion of a 72-bp exon (genomic exon 11) is common. We have shown that the presence of this 72-bp exon significantly enhances proliferation and invasion of immortalized prostate epithelial cells when compared with mRNAs with the same 5′ structure that lack this exon. Thus alternative splicing of the 5′ region and the coding exons of the T/E fusion gene, as well as the ratio of the various isoforms, can impact the biological activities of the fusion gene.

The T/E fusion gene can promote PCa invasion and to a lesser extent proliferation and decrease differentiation via several known pathways such as urokinase-type plasminogen activator (uPA), matrix metalloproteinase (MMP), and C-myc that have previously been implicated in PCa initiation and progression (5–8). Here, we report that NF-κB transcriptional activity is increased by T/E fusions. The NF-κB pathway activation is associated with aggressive clinical behavior in PCa (9). Phosphorylation of p65 has been shown by many groups to enhance p65 transcriptional activity (10). We have found that T/E fusion type isoforms enhance NF-κB transcriptional activity 3-fold in prostatic epithelial cells. Significantly increased phosphorylation of the NF-κB p65 subunit on Ser536 was observed in stably selected PNT1a cells expressing T/E fusion isoforms whereas VCaP cells with stable knockdown of the T/E fusion gene by shRNA show a marked decrease in p65 Ser536 phosphorylation. Using immunohistochemistry, we have found that p65 phospho-Ser536 is present in the majority of PCa tissues and its expression is associated with expression of the ERG protein. Further investigation revealed upregulation of several genes related to NF-κB pathway by the fusion gene, including Toll-like receptor 4 (TLR4). The TLR4 protein can activate the NF-κB pathway and increase p65 Ser536 phosphorylation when activated (11, 12). We verified upregulation of TLR-4 mRNA by T/E fusions and decreased p65 phospho-Ser536 in PNT1a cells expressing the T/E fusion was seen when TLR4 was knocked down. Thus, the NF-κB pathway is activated by phosphorylation via TLR4 activation in fusion gene expressing cells, linking the expression of this common fusion gene to a key pathway in PCa progression which can potentially be activated by bacterial lipopolysacharide (LPS) and/or tumor microenvironment proteins.

PNT1a cells were maintained in the RPMI with 10% FBS. HEK293T cells and Cos7 cells were maintained in DMEM medium with 10% FBS.

Transient transfection was conducted in triplicate in 24-well plates as described previously (13). The NF-κB luciferase reporter vector was obtained from Stratagene (Cat 219077, PathDetect, Stratagene). Luciferase activity was determined on triplicate samples and each experiment was repeated at least 3 times.

VCaP Cells (1.0 × 10⁵) of each cell line were trypsinized and plated in 6-well dishes in complete medium. The next day, medium containing 20 μmol/L PS1145 (Sigma Aldrich, P6624) were added to the wells in treatment group. Cells were trypsinized and counted using a Coulter counter at different time points in triplicate. The experiment was repeated twice.

Anti-p65 and anti-phos-p65 Ser536 (both from Cell Signaling Technology) were used at 1:1,000 dilution for Western blotting using procedures described previously (13). Anti-β-actin control was performed as described previously (13).

Immunohistochemistry of VCaP orthotopic tumors was performed using a rabbit polyclonal p65 anti-phospho-Ser536 from Abgent Inc. (AP3178a) as described previously (14). Slides were scanned and photographed using a Nikon Eclipse E400 microscope connected with Nuance Multispectral Imaging System at 200× magnification with 3.3 megapixel resolution. Images were saved as JPEG files with 6 to 8 images were taken for each slide, covering the entire tumor area. All staining signals are localized to cell nuclei. The numerical value for percent stained (PS) is determined by using Image software (http://rsb.info.nih.gov/ij/).

Cos7 cells were cultured in 4-well chamber slides (Lab-Tek) and transfected with NF-κB p65 along with T/E fusions type III+72 or VI+72, which have the V5 tag (4), or control vector (pCMV-Tag2B). Cells were fixed 48 hours after transfection with 4% formaldehyde for 15 minutes and incubated with 0.5% bovine serum albumin for 30 minutes. Antibodies (anti-V5 from Invitrogen, anti-p65, and anti-phospho-Ser536 from Cell Signaling) were applied to cells overnight at 4°C. Fluorescein isothiocyanate (FITC)-conjugated goat anti-mouse secondary antibody (1:500 dilution) was used for detection of fusion proteins, and tetramethyl rhodamine isothiocyanate (TRITC-conjugated goat anti-rabbit antibody (1:500 dilution) was used for total p65 or phospho-Ser536 p65 staining. Both secondary antibodies were purchased from Sigma-Aldrich, Inc. For nuclear staining, 1 mL of mounting solution was applied with DAPI (Vectashield, Vector) following manufacturer's instructions. Slides were mounted and observed under Nikon Eclipse E400 microscope connected with Nuance Multispectral Imaging System. Cells with positive fluorescence staining in 10 random fields at 200× magnification field were counted. This experiment was repeated twice.

PCa tissue microarrays have been described previously (15–18). Immunohistochemistry was performed as described above for the VCaP tumors. Arrays stained with anti-phospho-Ser536 were scanned and staining quantitated using a multiplicative staining index of intensity (0–3) and extent (0–3) of staining yielding a 10-point staining index (0–9) as described in previous publications (15–18). The arrays were also stained with anti-ERG antibody as described by Park and colleagues (19). Cancer nuclei were scored as positive or negative and cancers were considered positive if ERG staining was seen in cancer cells in any tissue core. In almost all cases in which positive staining was seen it involved the majority of cancer cells in all cores. No staining of benign epithelial nuclei was identified although strong staining of endothelial cells was noted as described previously (19) which was useful as an internal positive control.

Human NF-κB Signaling Pathway PCR Array from SA Biosciences was used to analyze cDNAs from PNT1a cells overexpressing T/E fusion type III+72, type VI+72 or vector control following the manufacturer's protocol. Individual PCR primer pair mixes for gene F2R, FOS, TLR3, TLR4, and CCL2 were purchased from SA Biosciences and were used for quantitative reverse transcriptase (RT)-PCR as described previously (20). This experiment was repeated twice.

Two shRNAs targeting TLR4 were purchased from Open Biosystems, [V2LHS_171352 (shRNAa) and V2LHS_171350 (shRNAb)]. Negative control shRNA was also purchased from Open Biosystems. Cells were transfected with shRNA plasmids against TLR4 or negative control shRNA as described previously (21). TLR4 mRNA levels were evaluated by quantitative RT-PCR as described above. This experiment was repeated twice.

To explore mechanisms by which T/E fusions affect PCa initiation and progression, we performed a series of luciferase reporter assays with a variety of reporter constructs and T/E fusion gene constructs (data not shown). Among the promoters tested, NF-κB transcriptional activities were the most highly upregulated. Briefly, we transiently transfected a composite NF-κB promoter construct linked to a luciferase reporter gene into the immortalized normal prostate epithelial PNT1a cell line with the plasmids expressing the III+72 fusion gene isoform, the VI+72 isoform, control empty vector plasmid or VI+72 deletion constructs missing key ERG domains. The type III+72 and VI+72 isoforms enhanced NF-κB transcription more than 2-fold and approximately 3-fold, respectively, compared with the control plasmid at 24 hours after transfection (Fig. 1) and this difference was statistically significant from control for both fusion gene isoforms (P < 0.02; t test). Similar results were seen at 8 to 72 hours after transfection (data not shown). Control transfections with VI+72 expression constructs with deletion of the ETS or CAE domain had activities that were even lower than control (Fig. 1), suggesting a possible dominant-negative activity directed against basal ERG expressed in PNT1a cells. Similar results were seen in 293T cells, although the induction was more robust (Supplementary Fig. S1).

Phosphorylation of p65 has been shown to enhance p65 transcriptional activity and phosphorylation at Ser536 regulates NF-κB p65 activation, nuclear localization, protein–protein interactions, and transcriptional activity (10, 22, 23). We therefore examined phosphorylation of p65 at Ser536 in stably selected PNT1a cells expressing T/E fusion gene isoforms. Both fusion gene expressing cell lines displayed significantly increased phosphorylation of the NF-κB p65 subunit on Ser536 (Fig. 2A) with unchanged total p65 protein. Of note, the VI+72 cell line had a higher level of p65 phospho-Ser536 than cells expressing the III+72 isoform. VCaP cells with stable knockdown of the T/E fusion gene by ShRNA that we have generated previously (see ref. 4) showed a marked decrease in levels of p65 phospho-Ser536. In previous studies, we used these shRNA and vector control VCaP cell lines for tumor progression studies following orthotopic injection and shown decreased tumor progression in shRNA expressing cells (4). We performed immunohistochemistry on 9 VCaP orthotopic tumors with stable knockdown of the fusion gene versus 10 vector controls using anti-p65 phospho-Ser536 antibody and quantitated percentage of cells with positive nuclear staining in each tumor. The tumors with stable T/E fusion knockdown had a significantly lower p65 phospho-Ser536 staining percentage by image analysis (36.9% vs. 48.8%; P < 0.002; t test). All staining was nuclear, with no cytoplasmic staining. Of note, VCaP cells, the only PCa cell line that expresses the T/E fusion gene, is also the only PCa cell line that expresses significant levels of Ser536 phosphorylated p65 (Fig. 2B). Therefore, T/E fusions enhance NF-κB activities through increased phosphorylation at the p65 Ser536 site.

We next determined the localization of T/E fusions and their regulation of p65 phosphorylation using fluorescence labeling techniques. The inactive form of NF-κB is localized in the cytoplasm and mainly consists of multiple subunits including the DNA-binding p50 and p65 subunits and an inhibitory subunit Iκβ, which is bound to p65 and masks the nuclear localization sequence and its release initiates the activation of NF-κB and its subsequent translocation to the nucleus. Cos7 cells were transfected with NF-κB p65 along with T/E fusions type III+72 or VI+72, which have a V5 tag (4), or control vector (pCMV-Tag2B). FITC-conjugated goat anti-mouse secondary antibody was used for detection of fusion proteins, and TRITC-conjugated goat anti-rabbit antibody was used for total p65 or phospho-Ser536 p65 staining. As expected, T/E fusion isoforms were seen exclusively in the nucleus with similar total expression level for each fusion type (data not shown). A similar number of cells with total nuclear p65 staining was observed in both fusion groups and the control group (Fig. 2C). A significantly higher number of p65 phospho-Ser536 expressing cells were found following cotransfection with fusion gene expressing constructs compared with the control group (Fig. 2D).

To determine whether NF-κB p65 Ser536 phosphorylation is relevant to human PCa, we carried out an immunohistochemical analysis of 371 clinically localized PCas using tissue microarrays. Expression was quantitated as described previously (15–18, 24–27) based on a multiplicative index of the average staining intensity (0–3) and extent of staining (1–3) in the cores yielding a 10-point staining index (0–9). Examples of strong (index 9), moderate (index 6), weak (index 3), and no (index 0) staining are shown in Figure 3A. The mean staining index was 4.5 and 85% of cancers showed some staining (i.e., >0). NF-κB p65 Ser536 phosphorylation was weakly correlated with Gleason score (r² = 0.122; P = 0.02) but not with other pathologic variables. Compared with cancers with no expression, cancers with p65 phospho-Ser536 had a significantly decreased time to recurrence (P = 0.005; log rank) as shown in the Kaplan–Meier plot in Figure 3B. Cox regression was significant by univariate (HR = 3.83, 95% CI: 1.41–10.43; P = 0.0086) and multivariate (HR = 3.64, 95% CI: 1.13–11.73; P = 0.03) analysis. Thus, the presence p65 Ser536 phosphorylation is common in PCa and influences clinical outcome.

We next explored the association of NF-κB p65 Ser536 phosphorylation expression in PCa with the expression of ERG using the protocol described by Park and colleagues (19) to detect ERG expression by immunohistochemistry in the same tissue microarray analyzed for p65 phospho-Ser536 expression. The presence of any level of ERG expression in tumor cells as detected by this antibody is highly correlated with the presence of the T/E fusion gene (19). We therefore dichotomized staining into positive and negative for ERG. Overall, 364 of 371 cancers informative for p65 phospho-Ser536 staining could also be evaluated for ERG expression. In the vast majority of cases, ERG staining was strong and in all cases was confined to cancer nuclei, with no staining of benign epithelium. Strong staining of endothelial cells was noted as described previously (19), which represents expression of endogenous ERG in these cells. A total of 164 cases were positive, representing 45% of evaluable cases, consistent with prior reported rates of T/E fusion gene expression in surgically treated PCa. ERG-positive cancers tended to have a shorter time to recurrence following radical prostatectomy (Supplementary Fig. S2), but the difference between ERG-positive and -negative cases was not statistically significant (P = .312; log rank). We then compared the expression index for p65 phospho-Ser536 in fusion gene–positive and fusion gene–negative cases. The expression index was significantly higher in ERG expressing PCa (mean = 5.29, median = 6.0) versus ERG-negative PCa (mean = 3.98, median = 3.0) by Mann–Whitney test (P < 0.001). Of note, there was no significant difference in the matched benign prostate tissues from ERG-positive and -negative cancers in p65 phospho-Ser536 expression. We also examined the fraction of cases that were ERG positive at various levels of p65 phospho-Ser536 expression. As seen in Figure 3C, the percentage of ERG-positive cases was only 17% in cases with no p65 phospho-Ser536 staining (staining index 0) and this increased to 60% in cases with strong staining (index 7–9). The differences seen were highly statistically significant (P < 0.001, χ²). To confirm the reproducibility of the NF-κB phospho-Ser536 scoring system when grouped in this manner, a second observer scored 270 individual tissue cores. Only 4 of 270 cores had a difference in scoring that resulted in change from one scoring group to another, indicating that this scoring system is highly reproducible. Overall, the data strongly support the concept that ERG can significantly enhance phosphorylation of p65 at Ser536 in vivo, although other factors clearly play a role because ERG-negative cases do have phosphorylation at this site, although at significantly lower levels. We then analyzed recurrence-free survival on 4 groups of cases grouped by p65 phospho-Ser536 and ERG expression status. (Supplementary Fig. S3). The ERG-positive and -negative cases that did not express p65 phospho-Ser536 were not statistically significantly different than each other. Among the cases expressing phospho-Ser536, the ERG-positive cases tended to have earlier recurrence than the ERG-negative cases, although this difference was not statistically significant (P = 0.34, log rank). The phospho-Ser536–positive/ERG-positive cases recurred earlier than both groups of phospho-Ser536 negative cases (P = 0.047 and 0.025, log rank). These data indicate that NF-κB Ser536 phosphorylation as a result of ERG expression impacts disease aggressiveness.

To confirm the importance of the NF-κB pathway in ERG expressing PCa cell lines, we treated VCaP cells with PS1145, an inhibitor of IKK (28). As shown in Figure 4, PS1145 significantly inhibited VCaP proliferation (P < 0.001, day 7, t test), confirming that NF-κB pathway plays an important role in proliferation in PCa cells expressing the T/E fusion gene.

To further explore the activation of the NF-κB pathway by the fusion isoforms, we used a SuperArray NF-κB PCR Array to analyze RNAs from vector control PNT1a cells and stably transfected PNT1a cells expressing the III+72 or VI+72 isoforms. Genes which were upregulated by 3-fold or more in one or both fusion gene expressing cells lines were then confirmed using standard quantitative RT-PCR. Results are shown in Figure 5. The five upregulated genes meeting this criterion were F2R (thrombin receptor; PAR1), TLR3, TLR4, FOS, and CCL2. The most significantly upregulated gene, particularly by the type VI+72 isoform, was TLR4.

Because the overexpression of T/E fusion isoforms can upregulate TLR4 expression and there is evidence that TLR4 can increase phosphorylation of p65 Ser536 when activated (11,12), we hypothesized that NF-κB transcription activated by T/E fusions is mediated through TLR4 pathway. Therefore, we carried out a knockdown experiment to determine whether p65 phospho-Ser536 level can be affected by decreased TLR4 expression in fusion gene expressing cells. We tested 2 ShRNA plasmids targeting TLR4 in PNT1a cells expressing the type VI+72 isoform, which showed the highest TLR4 expression at RNA level, by transient transfection. The knockdown efficiencies for the 2 shRNA plasmids are 53% and 40%, respectively, by quantitative RT-PCR (Fig. 6A). Dramatically decreased p65 phospho-Ser536 was seen in both transfectants by Western blot (Fig. 6B). Therefore, our in vitro data support our hypothesis that NF-κB p65 phospho-Ser536 upregulation by T/E fusion isoforms is mediated at least in part through TLR4.

We report here for the first time that T/E fusion isoforms can activate NF-κB transcriptional activity through increased p65 Ser536 phosphorylation. Clinical and preclinical observations have shown that NF-κB plays an important role in PCa growth, survival, angiogenesis, tumorigenesis, and metastatic progression (29–33). To date, investigations in PCa have focused on expression and translocation of p50/p65 subunits. Our data indicate that p65 phosphorylation also plays a role in modulating PCa progression and is increased in fusion gene expressing cells. Furthermore, our data indicate that TLR4 plays an important role in increasing p65 Ser536 phosphorylation in fusion gene expressing cells. Of note, ETS factor binding sites in the TLR4 promoter may play an important role of TLR4 mRNA expression (34), raising the possibility that TLR4 may be a direct target of the T/E fusion gene.

TLR4 is well known for its role in immune cells, in particular, its significant role in response to LPS produced by Gram (−) bacteria (11). DU145 and PC3 PCa cell lines also express TLR4 (35, 36). Knockdown of TLR4 in PC3 cells with siRNA significantly decreased invasion, survival, and tumorigenicity (35). Finally, it has been shown that specific sequence variants of TLR4 are associated with risk of PCa (37). TLR4 is a potent regulator of NF-κB signaling and has been shown to increase p65 Ser536 phosphorylation when activated (11, 12). It should be noted that although TLR4 is activated by LPS, it can also be activated by endogenous ligands, such as hyaluronic acid, fibronectin, and heparin sulfate (38), which are abundant in the tumor microenvironment. It has been suggested that activation of TLR4 by LPS released from infection of the prostate by Gram (−) bacteria may promote tumorigenicity (39). Symptomatic prostatitis occurs in approximately 10% of men whereas the prevalence of asymptomatic prostatitis in unknown. There is a correlation between inflammation and the detection of bacterial RNA in the PCa tissues (40). Whether endogenous ligands or LPS or both activate TLR4 in fusion gene expressing cells is an important question. If LPS does indeed play a role, this is clinically important in that it indicates that a potentially treatable infectious process may promote PCa initiation and/or progression.

TLR3 is also increased in response to fusion gene expression. TLR3 is expressed by both LNCaP and PC3 cell lines (41). TLR3 responds to dsRNA or poly I:C (12). Of note, activation of TLR3 can activate NF-κB and MAPK but can also induce apoptosis and has been shown to do so in LNCaP cells (41). Thus, it is possible that the observed induction of TLR3 may be exploited therapeutically in PCa if indeed TLR3 activation induces apoptosis rather than tumor promotion. Such direct antitumor effects would synergize with the known ability of dsRNA to promote antitumor immunity (42).

PAR1 (thrombin receptor) is a G-protein coupled receptor that has previously been shown to be overexpressed in human PCa (43,44) and is among the genes we found to be upregulated. Of note, PAR1 is strongly expressed and biologically active in VCaP cells (44). PAR1 is activated by thrombin and Factor 10a. The uPA pathway and MMPs (upregulated by the T/E fusion gene) can enhance PAR1 activation by enhancing thrombin formation in the tumor microenvironment. PAR1 can activate NF-κB in PCa cells (43) as well other cell types. Interestingly, a functional ETS element has been shown to be present at position 506 in the PAR1 promoter (45), raising the possibility that PAR1 may be direct transcriptional target of the T/E fusion gene protein.

CCL2 (MCP-1) is a chemokine which is a potent regulator of PCa cell migration and proliferation and acts, at least in part, via activation of the PI3 kinase/AKT pathway (46) and is a well-known direct transcriptional target of NF-κB (47). CCL2 is expressed by endothelial cells within the tumor microenvironment but can also be expressed by tumor cells directly (48). Of note, VCaP cells express CCL2 and secrete it into the media (48). Furthermore, the CCL2 receptor, CCR2, is expressed by PCa cells (48). CCL2 has been shown to significantly enhance growth of PCa bone metastases, the most common site of PCa metastasis and may act synergistically with PAR1 (see above) to promote migration (49).

Finally, we noted a significant upregulation of FOS by the T/E fusion gene. c-JUN is upregulated by the fusion gene and c-JUN can bind c-FOS. It is known that AP-1 can synergize with NF-κB to promote expression of specific target genes including proangiogenic genes such as interleukin (IL)-8 and VEGF (50). Thus, increased FOS expression may enhance the tumor promoting activities of the NF-κB pathway.

In summary, our study has identified a unique mechanism by which T/E fusion genes contribute to PCa tumorigenesis by upregulating the NF-κB pathway. We hypothesize that in a subset of PCa, T/E fusions can upregulate TLR4, and then activated TLR4 increases NF-κB transcriptional activities, resulting in expression of genes favoring tumor cell growth and invasion. This process can be enhanced by exposure of fusion gene expressing cells to Gram (−) bacteria, which releases LPS, and/or exposure to endogenous proteins in the tumor microenvironment. These findings have implications for prevention and treatment of PCa by treatment Gram (−) prostatic infections and targeting the NF-κB pathway.

No potential conflicts of interest were disclosed.

We gratefully acknowledge the technical assistance of DeeAnne Killeen with immunohistochemistry studies and Anna Frolov with statistical analysis.

This work was supported by grants from the Department of Defense Prostate Cancer Research program (DOD W81XWH-08-1-0055; J. Wang), from the National Cancer Institute to the Baylor Prostate Cancer SPORE (P50CA058204), and from the University of Michigan Prostate Cancer SPORE (P50CA069568), the Department of Veterans Affairs Merit Review program (M. Ittmann), and by the use of the facilities of the Michael E. DeBakey VAMC.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.