Ectopic fibroblast growth factor receptor 1 promotes inflammation by promoting nuclear factor-κB signaling in prostate cancer cells

Initiation of expression of fibroblast growth factor receptor 1 (FGFR1) concurrent with loss of FGFR2 expression is a well-documented event in the progression of prostate cancer (PCa). Although it is known that some FGFR isoforms confer advantages in cell proliferation and survival, the mechanism by which the subversion of different FGFR isoforms contributes to PCa progression is incompletely understood. Here, we report that fibroblast growth factor (FGF) promotes NF-κB signaling in PCa cells and that this increase is associated with FGFR1 expression. Disruption of FGFR1 kinase activity abrogated both FGF activity and NF-κB signaling in PCa cells. Of note, the three common signaling pathways downstream of FGFR1 kinase, extracellular signal–regulated kinase 1/2 (ERK1/2), phosphoinositide 3-kinase (PI3K/AKT), and phosphoinositide phospholipase Cγ (PLCγ), were not required for FGF-mediated NF-κB signaling. Instead, transforming growth factor β–activating kinase 1 (TAK1), a central regulator of the NF-κB pathway, was required for FGFR1 to stimulate NF-κB signaling. Moreover, we found that FGFR1 promotes NF-κB signaling in PCa cells by reducing TAK1 degradation and thereby supporting sustained NF-κB activation. Consistently, Fgfr1 ablation in the transgenic adenocarcinoma of the mouse prostate (TRAMP) model reduced inflammation in the tumor microenvironment. In contrast, activation of the FGFR1 kinase in the juxtaposition of chemical-induced dimerization (CID) and kinase 1 (JOCK1) mouse model increased inflammation. As inflammation plays an important role in PCa initiation and progression, these findings suggest that ectopically expressed FGFR1 promotes PCa progression, at least in part, by increasing inflammation in the tumor microenvironment.

Cancer-related inflammation is one of the hallmark changes found during cancer progression (1,2). The tumor microenvironment is home to multiple types of inflammatory immune cells, including CD4 ϩ T cells, macrophages, and neutrophils, which facilitate cancer progression and recurrence (3)(4)(5)(6). Prostate cancer (PCa), 4 the second most frequently diagnosed cancer in men worldwide (7), is no exception. Emerging evidence demonstrates that chronic inflammation promotes the development of PCa to an advanced metastatic disease (8,9) that is castration-resistant (10,11).
Tumor cells produce multiple cytokines and chemokines to attract infiltrating lymphocytes to the tumor microenvironment. Among these are the upstream regulators of nuclear factor (NF)-B, which was first discovered as a transcription factor that promotes expression of the immunoglobulin chain by binding with its enhancer during the maturation of B lymph cells (12). NF-B plays important roles in the cytokine/chemokine-mediated immune response (8,9,13,14). The NF-B transcription factor family consists of five proteins, p65 (RelA), RelB, c-Rel, p105/p50, and p100/p52, which associate with each other to form distinct homo-and heterodimeric complexes (15). In its resting state, NF-B dimers are inactivated by interacting with the inhibitor of B (IB), preventing NF-B nuclear translocation (12). Many cytokines and chemokines, including tumor necrosis factor ␣ (TNF␣) and interleukin-1, activate NF-B. Upon activation by TNF␣, the TNF␣ receptor recruits TNF receptor-associated factor 2/5 (TRAF2/5) and transforming growth factor ␤-activated kinase 1 (TAK1) (16). Activated TAK1 then phosphorylates and activates the inhibitor of B kinase ␣/␤ (IKK␣/␤), leading to the phosphorylation and degradation of IB␣. The released NF-B is then translocated into the nucleus and initiates target gene transcription (17).
The NF-B pathway is constitutively activated in a significant portion of human PCa. This dysregulation correlates with the castration resistance of PCa (13). Consistent with clinical observations, the NF-B pathway is also highly activated in androgen-independent prostate cancer (PC)-3 and Duke University (DU)145 cells, but not in androgen-dependent lymph node carcinoma of the prostate (LNCaP) cells (14). Moreover, NF-B signaling promotes cytokine expression and increases tumor-infiltrating macrophages that facilitate PCa survival, metastasis, and angiogenesis (18,19). However, the mechanism by which the NF-B pathway is dysregulated in PCa cells is not understood.
The fibroblast growth factor (FGF) family consists of 18 tyrosine kinase receptor-binding ligands regulating a broad spectrum of cellular processes. FGFs bind and activate the FGF receptor (FGFR), transmembrane tyrosine kinases encoded by four highly homologous genes, denoted Fgfr1, Fgfr2, Fgfr3, and Fgfr4 (20). Ectopic expression and activation of the FGF signaling axis is associated with many diseases, including developmental disorders and cancer (21). Loss of resident FGFR2 and acquisition of ectopic FGFR1 is a common event in PCa progression (22). However, how this conversion of FGFR isoform expression contributes to PCa progression is not fully understood. In addition, the cross-talk between the FGF and NF-B signaling pathways remains controversial. In some reports, FGF signaling positively regulates the NF-B pathway (23)(24)(25)(26)(27)(28)(29), whereas in others, FGF signaling negatively regulates the NF-B pathway (30 -32). In addition, the mechanism by which FGFR regulates the NF-B pathway is unclear, although downstream signaling pathways of FGFR kinases, such as the extracellular signal-regulated protein kinases (ERKs), phosphatidylinositol 3-kinase (PI3K)/AKT, or phospholipase C (PLC)-␥ pathways, are known to promote activation of the NF-B pathway.
In this report, we demonstrated that FGF augmented NF-B signaling in FGFR1-expressing, but not in FGFR2-expressing, PCa cells; that FGFR1 augmented NF-B signaling via tyrosine phosphorylation, thereby stabilizing TAK1 and sustaining NF-B signaling; and that inflammation in the PCa microenvironment was associated with ectopically expressed FGFR1 kinase. Inflammation has long been known to promote cancer initiation and progression, and our results suggest that FGFR1 promotes inflammation in a receptor isoform-specific manner, which, at least in part, accounts for its activity to promote PCa progression.

FGF1 promotes NF-B signaling in DU145 cells that highly express FGFR1, but not in LNCaP cells that do not express FGFR1
Although the acquisition of ectopic FGFR1 and elevated inflammation are characteristic events in PCa progression, the cross-talk between FGF and NF-B signaling remains controversial. To determine the presence or absence of cross-talk between ectopic FGFR1 signaling and the NF-B pathway in PCa cells, DU145 cells that highly expressed FGFR1 and LNCaP cells that highly expressed FGFR2 were treated with FGF1 (Fig.  1A). Western blotting analyses revealed that, although FGF1 induced strong phosphorylation of fibroblast growth factor receptor substrate 2␣ (FRS2␣) and ERK1/2 in both DU145 and LNCaP cells, a time-dependent phosphorylation of IKK␣ and IKK␤ was observed only in DU145 cells (Fig. 1B). In the NF-B pathway, phosphorylation activates IKK␣/␤, which in turn phosphorylates IB␣, leading to degradation of IB␣ and, therefore, activation of NF-B. Treatment of DU145 cells with FGF1 consistently increased the phosphorylation and reduced the abundance of IB␣. Interestingly, FGF1 also induced phosphorylation of p65 in DU145 cells (Fig. 1B), further indicating activation of the NF-B pathway. In contrast, FGF1 failed to induce phosphorylation of IKK␣, IKK␤, IB␣, or p65, as well as to reduce IB␣ abundance in LNCaP cells (Fig. 1C). Together with the finding that DU145, but not LNCaP, cells highly expressed FGFR1, these results suggest that the cross-talk between the FGF and NF-B pathways is FGFR1-dependent.

Disruption of the FGFR1 pathway compromises NF-B signaling in PCa cells
To determine whether FGFR kinase activity was required for FGF to promote NF-B signaling, DU145 cells were incubated with the FGFR kinase inhibitor, AZD4547, before FGF treatment. Western blotting analyses revealed that AZD4547 blunted the activity of TNF␣, evidenced by reduced phosphorylation of IKK␣/␤ and p65, and increased IB␣ abundance FGF promotes NF-B signaling in prostate cancer ( Fig. 2A). Furthermore, real-time RT-PCR analyses revealed that AZD4547 suppressed the activity of TNF␣ and FGF, reducing expression of TNF␣ and the chemokine CXC motif ligand 1 (CXCL-1), both of which are downstream targets of the NF-B pathway (Fig. 2B). To determine whether FGFR1 was required for FGF to promote NF-B signaling, the CRISPR/ Cas9 gene editing method was used to ablate Fgfr1 in DU145 cells (designated DU145 ⌬R1 ). Unlike control DU145 cells, where TNF␣ induced nuclear translocation of p65 within 15 min, DU145 ⌬R1 cells failed to respond to TNF␣ with respect to p65 nuclear translocation (Fig. 3A). Immunostaining further demonstrated that TNF␣ induced p65 translocation in control DU145 cells, but not DU145 ⌬R1 cells (Fig. 3B). Forced expression of FGFR1 restored TNF␣-induced p65 translocation in DU145 ⌬R1 cells. Similar to treatment with AZD4547, ablation of Fgfr1 compromised the TNF␣-induced activation of NF-B and abrogated the ability of FGF1 to augment TNF␣ signaling in DU145 cells (Fig. 3C). Together, these results indicate that FGFR kinase activity is required for cross-talk between the FGF and NF-B signaling pathways.

TAK1 is required for FGFR1 to augment NF-B signaling
The ERK1/2, PI3K/AKT, and PLC␥ pathways are the three common pathways downstream of FGFR. To determine whether these three pathways were required for FGF to augment NF-B signaling, mouse embryonic fibroblasts (MEFs) were incubated with the ERK, AKT, or PLC␥ inhibitors SL327, LY294002, or U73122, respectively, 1 h before FGF and TNF␣ treatment. Western blotting analyses revealed that suppression of either ERK, AKT, or PLC␥ did not affect the ability of FGF1 to augment NF-B signaling (Fig. 4A), despite the fact that the inhibitors suppressed phosphorylation of ERK and AKT or Ca 2ϩ influx (data not shown), respectively. These results, indicating that inhibition of PI3K with LY294002 did not affect the phosphorylation of IKK␣/␤ in MEFs, differ from previous results indicating that FGF1 activates NF-B via the AKT pathway in breast cancer cells (28). This is likely another example of cell context-dependent signaling of the FGF pathway. However, the underlying mechanism remains to be determined.
TAK1 is a member of the mitogen-activated kinase kinase kinase family. TAK1 can be activated by multiple cytokines or by coexpression with the TAK1-binding protein (TAB1) (34). Upon activation, TAK1 phosphorylates IKK␣/␤, thereby activating the canonic NF-B pathway. Whether TAK1 is required for the FGF pathway to cross-talk with NF-B is not fully clear, although one report found that FGFR interacts with TAK1 in multiple myeloma and bladder cancer cells (23). To determine whether the FGF pathway enhanced NF-B signaling via activation of TAK1, MEFs were treated with the TAK1 inhibitor, (5Z)-7-oxozeaenol. Inhibition of TAK1 abrogated the FGF1induced phosphorylation of IKK␣/␤ (Fig. 4A). These data suggest that TAK1 kinase activity is required for FGF to promote NF-B activity in MEFs.
Furthermore, to determine whether TAK1 was required for FGF to augment NF-B signaling, MEFs bearing null Tak1 alleles (MEF ⌬TAK1 ) were treated with FGF1 and TNF␣. Western blot analysis revealed that both FGF1 and TNF␣ failed to induce phosphorylation of IKK␣/␤ in MEF ⌬TAK1 cells (Fig. 4B). Forced expression of TAK1 in MEF ⌬TAK1 cells by stable transfection restored TNF␣-induced IKK␣/␤ phosphorylation as well as the ability of FGF1 to enhance TNF␣-induced phosphorylation of IKK␣/␤ (Fig. 4B).

FGFR1 forms a complex with TAK1 and leads to TAK1 tyrosine phosphorylation
To determine the mechanism by which FGFR1 cross-talked with the NF-B pathway, FLAG-tagged FGFR1 was co-expressed with MYC-or HA-tagged components of the TNF␣ receptor complex in human embryonic kidney (HEK)293 cells. Immunoprecipitation with anti-MYC or anti-HA antibodies was employed to pull down FGFR1 that bound to the MYC-or HA-tagged proteins. The resulting complexes were then subjected to Western blotting analyses (Fig. 6A); FGFR1 was pulled down by TAK1 and the TAK1-TAB1 complex, but not by other components of the TNF␣ receptor complex. These results indicate the presence of an interaction between FGFR1 and TAK1.

FGF promotes NF-B signaling in prostate cancer
To determine the structure domains of TAK1 required for interaction with FGFR1, either MYC-tagged full-length TAK1, kinase domain, or C-terminal sequence downstream of the kinase domain was co-expressed with FLAG-tagged FGFR1 or a constitutively active mutant of FGFR1 (caFGFR1) in HEK293 cells. Co-immunoprecipitation analysis demonstrated that only the full-length TAK1 and the kinase domain pulled down FGFR1 (Fig. 6B), suggesting that TAK1 interacted with FGFR1 via its kinase domain. In addition, the relative amount of caFGFR1 pulled down was greater than that of WT FGFR1, suggesting that kinase activity enhanced the interaction between FGFR and TAK1.
Further in silico screening identified four potential tyrosine phosphorylation sites in the kinase domain of TAK1 (Fig. 6C). Site-directed mutagenesis was utilized (Tyr 3 Phe, singly and in combination) to determine the phosphorylation status of these tyrosine residues and whether the phosphorylation was FGFR1-dependent. Western blotting analyses revealed that TAK1 was tyrosine-phosphorylated only when co-expressed with caFGFR1 in 293 cells (Fig. 6D), indicating that the phosphorylation was FGFR1 kinase-dependent. Mutations of each individual tyrosine residue alone, as well as two tyrosine residues together (2Fa and 2Fb, as indicated) only partially reduced phosphorylation. However, no tyrosine phosphorylation was detected with the 4F mutant (all four tyrosine residues substituted with phenylalanine) (Fig. 6E). The results indicate that all four tyrosine residues were phosphorylated.

Elimination of the four tyrosine phosphorylation sites of TAK1 accelerates its lysosome-mediated degradation
TAK1 is activated through binding to TAB1. However, no significant difference was observed in TAB1 binding between WT TAK1 and the TAK1 4F mutant (data not shown). Lys-48 ubiquitination-mediated lysosome degradation is involved in inactivation of TAK1 activity (35,36), and Lys-63 ubiquitination on Lys-158 is involved in TAK1 activation (36 -38). Therefore, we investigated whether the ubiquitination of TAK1 was regulated by tyrosine phosphorylation. Interestingly, we found that the ubiquitination of TAK1 was increased in DU145 ⌬R1 cells when they were treated with lysosome inhibitor MG132 (Fig. 7A). Consistently, the 4F mutant also exhibited increased ubiquitination in DU145 cells in the presence of MG132. To

FGF promotes NF-B signaling in prostate cancer
determine whether tyrosine phosphorylation suppresses ubiquitin-mediated degradation of TAK1, both TAK1 WT and tyrosine phosphorylation site mutants were co-expressed with caFGFR1 in HEK293 cells. The cells were treated with cycloheximide to halt protein synthesis followed by the addition of the proteasome inhibitor, MG132, for the indicated times. Western blot analysis was then used to assess the abundance of WT and mutant TAK1 at different time points with or without treating with MG132 (Fig. 7, B and C). Quantitative analyses of TAK1 intensity revealed that the 4F mutant exhibited a faster down-regulation than other TAK1 constructs. Treating the cells with lysosome inhibitor, MG132, diminished the differ-ences in down-regulation, indicating that the 4F mutant was down-regulated faster than WT and other mutants by proteasome-mediated degradation. These results suggest that tyrosine phosphorylation suppresses degradation of TAK1 and therefore sustains NF-B signaling.

Ablation of Fgfr1 in mouse PCa reduces inflammation in the tumor microenvironment
To investigate how ectopic FGFR1 contributed to inflammation in the tumor microenvironment, sections of PCa derived from TRAMP mice with or without ablation of Fgfr1 alleles were immunostained with antibodies against cyclooxygenase

FGF promotes NF-B signaling in prostate cancer
(COX)-2 (expressed only in inflammatory tissues) or F4/80 (a macrophage-specific marker) (Fig. 8A). The results clearly demonstrated that ablation of Fgfr1 reduced the abundance of COX-2 and F4/80 in PCa. In addition, similar experiments were carried out with the juxtaposition of CID and kinase 1 (JOCK1) PCa mouse model, where PCa was induced through chemicalinduced dimerization of the FGFR1 kinase domain-FK506binding protein fusion protein (39). We found that activation of the fusion protein with chemical-induced dimerization also increased the staining of COX-2 and F4/80 in the prostate tissues (Fig. 8B). Together, these results demonstrate that activation of FGFR1 in prostate cells increases expression of COX-2 and infiltration of macrophages, both of which are indicative of inflammation, and therefore support the in vitro finding that FGFR1 promotes inflammation.

Discussion
Inflammation is a characteristic change in the tumor microenvironment, which promotes cancer initiation and progression. In addition, acquisition of ectopic FGFR1 expression is a common event during PCa progression. In this report, we demonstrated that ectopic FGFR1 promoted inflammation by preventing the proteasome-mediated degradation of TAK1 and therefore sustaining NF-B signaling. Furthermore, we found that ablation of Fgfr1 in mouse PCa reduced inflammation and infiltration of macrophages into the tumor microenvironment. Therefore, these results reveal a novel mechanism by which ectopic FGF signaling contributes to PCa progression.
AKT, ERK, and PLC␥ pathways are well characterized downstream signaling cascades in FGF signaling transduction. However, these are also common downstream pathways for multiple receptor tyrosine kinases, including epithelial growth factor, platelet-derived growth factor, and vascular endothelial growth factors. The mechanism by which FGFR1 elicits receptor-specific signals is poorly understood. Our data clearly demonstrate that cross-talk between FGF and NF-B signaling is independent of ERK, PI3K/AKT, and PLC␥ pathways and that FGFR1 directly interacts with TAK1. Interestingly, FGFR2, although highly homologous to FGFR1, did not interact with TAK1. 5 Together, these results reveal a novel pathway whereby

FGF promotes NF-B signaling in prostate cancer
elicits receptor isoform-specific signals. Inflammation and NF-B signaling are complex processes and can be regulated by multiple cell signaling pathways at multiple levels. In this report, we demonstrated that FGFR1 promoted NF-B by sustaining the stability of TAK1. However, we did not exclude the possibility that FGFR1 may also be promoting NF-B signaling via other mechanisms. Interestingly, our data reveal that FGF alone only weakly activates NF-B signaling. However, it augments TNF␣ activity, suggesting that FGF signaling provides another layer of controlling NF-B activation. DU145, MEF, and HEK293 cells may have distinct background TNF␣ and FGF signaling (autocrine); therefore, FGF alone exhibits a different effect on NF-B activation in these three cell lines.
TNF␣ receptor adaptor RIP, TRAF2, and TRAF6 proteins are required for TNF-induced Lys-63-linked polyubiquitination of TAK1 (40 -43). These polyubiquitin chains serve as a scaffold to recruit the TAK1 and IKK complexes and facilitate phosphorylation of IKK␤ by TAK1, leading to activation of NF-B. On the other hand, Lys-63-linked polyubiquitin chains can be disassembled by deubiquitination enzymes, including CYLD and A20, which terminates IKK activation (44). In addition, CYLD and A20 also induce Lys-48 -linked polyubiquitination of TAK1, which mediates TAK1 degradation. Whether RIP, TRAF2, and TRAF6 are involved in the complex of FGFR1 and TAK1 and whether FGFR1 regulates assembly and disassembly of Lys-63-or Lys-48 -linked polyubiquitin chains is unknown, and these remain as interesting questions. Furthermore, the mechanism by which FGFR1 induces tyrosine phosphorylation of TAK1 and thereby controls its ubiquitination and subsequent activation or proteasome-mediated degradation remains to be elucidated.
Chronic inflammation in the tumor microenvironment is a key contributor to tumor initiation and progression. Acquisition of ectopic FGFR1 concomitant with the loss of FGFR2 is a common event in PCa progression (22). Our data demonstrated that FGFR1, but not FGFR2, promoted NF-B signaling and, therefore, inflammation. This is consistent with the fact that normal prostate tissue has no detectable inflammation nor expressed resident FGFR2, whereas PCa has a highly inflammatory tumor microenvironment and ectopically expresses FGR1 (22,45). Our results further demonstrate that FGFR1 elicits isoform-specific signals to promote inflammation; the FGFR isoform-specific interaction between FGFR1 and TAK1 accounts for the FGFR isoform-specific cross-talk. However, the existence of additional mechanisms underlying the receptor isoform-specific interaction between FGF and NF-B signaling pathways remains to be elucidated.
In addition to promotion of cell proliferation and survival, we also demonstrated that FGFR1 promoted tumor angiogenesis and cross-talk with stroma cells (46,47). Furthermore, we found that ectopically expressed FGFR1 promoted tumor microenvironment remodeling, a promoter for PCa initiation and progression (47). In line with this, data in Fig. 8 clearly The cells were treated with MG132 for 24 h followed by cycloheximide to halt protein synthesis. The cells were lysed at the indicated time points, and the lysates were subjected to Western blotting analyses with anti-MYC antibodies (B). The intensity of MYC bands was analyzed with ImageJ, normalized with that of ␤-actin, and plotted (C). IB, immunoblot; IP, immunoprecipitation; Ub, ubiquitin; CHX, cycloheximide.

FGF promotes NF-B signaling in prostate cancer
demonstrated that ablation of Fgfr1 reduced inflammation in the TRAMP tumor. In contrast, activation of the FGFR1 transgene promoted inflammation in JOCK1 PCa. Overexpression of FGFR1 and ectopic FGF signaling has been found in Ͼ80% of PCa (21). Further efforts are needed to determine whether overexpression of FGFR1 is associated with elevated inflammation in human PCa. Multiple FGFR kinase inhibitors, including TKI258, have been used in clinical trials for PCa and other cancers (21). Whether these inhibitors suppress inflammation in the tumor microenvironment of PCa and whether the effects are associated with FGFR1 expression remain to be determined.
In conclusion, FGFR1 augments NF-B signaling by preventing the proteasome-mediated degradation of TAK1. Disruption of FGFR1 compromised NF-B signaling and reduced inflammation in the tumor microenvironment. As the acquisition of ectopic FGFR1 is a key event in PCa progression, inhibition of ectopic FGFR1 signaling may retard PCa progression through suppressing chronic inflammation, therefore providing a potential new strategy for controlling PCa progression.

Animals
Mice were housed under the Program of Animal Resources of the Institute of Biosciences and Technology in accordance with the principles and procedures of the Guide for the Care and Use of Laboratory Animals. All experimental procedures were approved by the Texas A&M University institutional animal care and use committee. Mice were bred and genotyped as described previously (48 -50).

FGF promotes NF-B signaling in prostate cancer
cells cultured overnight (1 ϫ 10 5 cells/well in 6-well plates) were transfected with 5 g of pcDNAZeo plasmid (Invitrogen) bearing the indicated cDNA constructs and 5 l of Lipofectamine (Invitrogen). The cells were then incubated at 37°C for 24 h before analysis.
The membranes were washed with TBS with Tween 20 (TBST) buffer to remove nonspecific antibodies and then incubated with horseradish peroxidase-conjugated rabbit antibody at room temperature for 1 h. The specifically bound antibodies were then visualized using ECL-Plus chemiluminescent reagents from Thermo Fisher Scientific. The films were scanned with a densitometer for quantitation.

Immunoprecipitation
Cells were lysed with RIPA buffer as described earlier. The cell lysates were incubated with anti-HA (1:1,000), anti-MYC (1:1,000), anti-TAB1 (1:1,000), or anti-FLAG (1:2,500) antibodies and protein A beads overnight at 4°C as described previously (51). The immunocomplexes were washed with RIPA buffer four times, boiled with 30 l of SDS sample buffer for 5 min, and then subjected to Western blot analysis with the indicated antibodies as described previously.

RNA expression
Total RNA was isolated from cells using TRIzol RNA isolation reagents from Life Technologies. The first-strand cDNAs were reverse transcribed from the RNA template using Super-Script III reverse transcriptase (Invitrogen) and random primers according to the manufacturer's protocols. Real-time PCR analyses were carried out using the Fast SYBR Green Master Mix (Life Technologies) as instructed by the manufacturer. The relative abundance of mRNA was calculated using the compar-ative threshold cycle method and normalized to ␤-actin as the internal control. The mean Ϯ S.D. among at least three individual experiments are shown.

Site-directed mutagenesis
The QuikChange Lightning multisite-directed mutagenesis kit (Agilent Technologies, Santa Clara, CA) was used for generating TAK1 mutant cDNAs. Primer sequences used to generate the mutants were as follows: Y124F, GTG CTG AAC CAT TGC CAT TTT ATA CTG CTG CCC ACG CAA; Y125F, GTG CTG AAC CAT TGC CAT ATT TTA CTG CTG CCC ACG CAA; Y124F/Y125F, GTG CTG AAC CAT TGC CAT TTT TTA CTG CTG CCC ACG CAA; Y143F, GTT CCC AAG GAG TGG CTT TTC TTC ACA GCA TGC AAC CCA; Y206F, TTT TTG AAG GTA GTA ATT TCA GTG AAA AAT GTG ACG TCT.

Gene ablation
The lentivirus-based CRISPR-Cas9 system was used to ablate Fgfr1 alleles in DU145 cells. The sequence of single guide RNA (sgRNA) was AACTTGTTCCGATGGTTATC. Two days after infection with the lentivirus, the virus-containing cells were selected by growing in medium containing 2 g/ml puromycin.

Protein stability assay
Stable transfectants of HEK293 cells expressing MYC-tagged TAK1 or TAK1 mutants were treated with cycloheximide for the indicated times. The abundance of MYC-tagged proteins was examined via Western blot analysis. The relative levels of specific bands were quantitated by ImageJ.

Immunostaining
PCa sections were rehydrated and stained with hematoxylin and immunostained as described previously (33). For p65 immunostaining, DU145 cells were fixed with 4% paraformaldehyde for 2 h at room temperature. The antigens were retrieved by boiling in citrate buffer (10 mM, pH 8.0) for 20 min. The rabbit anti-COX-2 antibody (1:1,000) was obtained from Cell Signaling Technology. Mouse anti-p65 (1:1,000) and rat anti-F4/80 (1:1,000) were obtained from Abcam. The ExtrAvidin peroxidase system (Sigma-Aldrich) and fluorescence-conjugated secondary antibodies (Invitrogen) were used to visualize specifically bound antibodies. For immunofluorescence staining, the nuclei were counterstained with TO-PRO-3 before observation under a confocal microscope (Zeiss LSM 510).

Statistical analysis
Statistical analysis was performed using a two-tailed t test, with significance set to p Ͻ 0.05. Error bars indicate S.D.