Oncogene (2002) 21, 5069 – 5080 ª 2002 Nature Publishing Group All rights reserved 0950 – 9232/02 $25.00 www.nature.com/onc
Regulation of FGF8 expression by the androgen receptor in human prostate cancer Vincent J Gnanapragasam1, Craig N Robson1, David E Neal1 and Hing Y Leung*,1 1
Prostate Research Group, School of Surgical Sciences, University of Newcastle-upon-Tyne, Framlington Place, Newcastle-uponTyne, NE2 4HH, UK
Fibroblast growth factor 8 (FGF8) has been shown to play a key role in prostate carcinogenesis. It was initially cloned as an androgen induced protein in mouse mammary cancer SC3 cells. In this study, we examined if FGF8 was also regulated by the androgen receptor in human prostate cancer. FGF8b protein expression in resected clinical prostate cancer correlated closely with expression of the androgen receptor (AR). In the androgen sensitive CWR22 prostate xenograft, we observed up-regulation of FGF8b immunoreactivity in testosterone supplemented mice while castration markedly reduced its signal. Furthermore, FGF8b protein expression in AR positive LNCaP cells was similarly enhanced by androgens. The proximal promoter of the human FGF8 gene was cloned into a luciferase reporter construct (FGF8.luc). FGF8.luc activity in AR positive LNCaP and SC3 cells was increased 2.5-fold by androgens. In AR negative DU145 cells, maximal induction of FGF8.luc required both cotransfection of the AR and the presence of androgens. The anti-androgen bicalutamide completely abolished AR mediated FGF8.luc induction. Deletion constructs from FGF8.luc have further defined an active promoter region and an androgen responsive region. Nucleotide analysis of this androgen responsive region has revealed putative androgen response elements. Finally, using ChIP assays we confirmed in vivo interaction between the AR and the androgen responsive region of the FGF8 promoter. Taken together these data provide first evidence that expression of the mitogen FGF8 in prostate cancer is, at least in part, regulated by the androgen receptor at the transcriptional level. Oncogene (2002) 21, 5069 – 5080. doi:10.1038/sj.onc. 1205663 Keywords: FGF8; prostate cancer; androgen receptor; gene promoter Introduction Prostate cancer is the second most common cause of cancer deaths in men in western countries (Greenlee et al., 2000; Merril et al., 1996). Given the ageing
*Correspondence: HY Leung; E-mail:
[email protected] Received 29 October 2001; revised 17 April 2002; accepted 10 May 2002
population, it is likely that prostate cancer will further increase as a clinical problem. The mainstay of therapy for advanced disease is androgen deprivation with monitoring for treatment response by measurement of prostate specific antigen (PSA). Thirty per cent of cancers do not initially respond to androgen deprivation and up to 70% of cases with initial good response will relapse after 2 – 3 years (Bower and Waxman, 1997). The molecular mechanism for the development of hormone refractory disease remains poorly understood. Recent studies have shown that the AR continues to be expressed in a proportion of hormone relapsed prostate cancer (Van der Kwast et al., 1991; De Vere White et al., 1997; Koivisto and Helin, 1999). In this context, gene expression regulated by the androgen receptor is of crucial significance. Fibroblast Growth Factor 8 (FGF8) was first identified from the culture media of SC3 mouse mammary carcinoma following treatment with androgens (Tanaka et al., 1992). It belongs to the family of FGFs that are known to be involved in foetal development, central nervous system development, angiogenesis and wound healing (Crossley et al., 1996; Lee et al., 1997). The human FGF8 gene is encoded on 10q24 and is phenotypically expressed as four isoforms in the human as compared to seven isoforms in the mouse (Gemel et al., 1996; Yoshiura et al., 1997). The difference among isoforms resides mainly in the amino terminus of the mature protein (36 kDa) with FGF8b being the most abundant and transforming isoform (Ghosh et al., 1996). The tumourigenic effect of FGF8 has been demonstrated by over-expression experiments (Kuohara et al., 1994; MacArthur et al., 1995; RudraGanguly et al., 1998). Generation of FGF8b transgenic mice has also been shown to induce mammary and salivary gland tumours as well as ovarian stromal hyperplasia (Daphna-Iken et al., 1998). FGF8 protein is over-expressed in human prostate and breast cancers (Tanaka et al., 1998; Wu et al., 1997) and previous work from our centre has shown that FGF8 mRNA is over-expressed in 60 – 70% of newly diagnosed prostate cancers (Leung et al., 1996). The level of expression also correlates with tumour stage, pathological grade and disease specific survival (Dorkin et al., 1999). Furthermore the receptors for FGF8, namely FGFR2IIIc, FGFR3IIIc and FGFR4 are known to be expressed in malignant prostate epithelium (Story et al., 1994; Ittman and Mansukhani, 1997).
Regulation of FGF8 expression in the prostate VJ Gnanapragasam et al
5070
The prostate gland is closely regulated by androgens acting via the AR during both development and in adulthood. The presence of androgen is a pre-requisite for the development of prostate cancer. Despite the initial observation of androgen inducible FGF8 expression in SC3 cells, the role of the AR in FGF8 expression in the human prostate has not been fully investigated. We therefore tested if the AR regulated FGF8 expression in prostate cancer, particularly at the transcriptional level.
hormone refractory cancers, we demonstrated FGF8b staining in 12 out of 15 cases (Figure 1c) and AR staining in 11 out of 15 cases (Figure 1d). Significantly the majority of cases that were AR positive were also FGF8b positive (10 out of 11). Two out of four hormone refractory cancers were negative for both FGF8b and AR immunoreactivity (Figure 1e and f respectively). FGF8b was expressed in human breast cancer (Figure 1g – positive control) but not in normal bronchus (Figure 1h – negative control) or sections incubated with no primary antibody (data not shown).
Results FGF8b is differentially expressed in clinical prostate cancer Using immunohistochemistry, FGF8b protein was found in all cases (20 out of 20) of primary untreated prostate cancer (Figure 1a) with varying levels of intensity, as has been previously reported (Leung et al., 1996; Tanaka et al., 1998). AR expression was also consistently found in these cases (Figure 1b). In
In vitro and in vivo FGF8b expression is associated with AR expression and is modulated by androgens To test the hypothesis that FGF8b expression is influenced by androgens, we used the CWR22 prostate xenograft model. Tumour growth and PSA expression has been shown to be exquisitely androgen-regulated in this model (Wainstein et al., 1994). All tumours were initially grown in untreated male nude mice and at the start of the study period, tumour sizes in each
Figure 1 FGF8b expression in clinical prostate tissue. (a) Section of untreated prostate cancer (pathology grade 8 – Gleason sum score) showing strong FGF8b signals. (b) Section of untreated prostate cancer with positive AR staining. (c and d) Serial sections of hormone refractory cancer with positive staining for FGF8b and AR respectively. (e and f) Serial sections of hormone refractory cancer with negative staining for FGF8b and AR respectively. (g) Section of breast cancer stained for FGF8b (positive control). (h) Section of bronchus stained for FGF8b (negative control) Oncogene
Regulation of FGF8 expression in the prostate VJ Gnanapragasam et al
5071
randomized group were very similar as detailed in Materials and methods. Tumour sizes changed significantly in response to treatment. In castrated mice, tumours had a mean size of 111.5+7.7 mm3. In testosterone-treated mice tumours were significantly larger (826.5+6.3 mm3) than those in castrated mice (P50.005, Fisher’s Exact Test). This confirmed that the model was androgen responsive in our hands. In addition we observed that PSA immunoreactivity in these tumours was reduced by castration and induced by testosterone treatment (Figure 2, lower panel). AR staining was reduced by castration but only mildly induced by testosterone treatment (Figure 2, middle panel). FGF8b immunoreactivity was reduced in tumours of castrated mice and up-regulated in tumours of testosterone treated mice (Figure 2, upper panel). To further confirm this latter finding, we used cultured AR positive LNCaP cells in which proliferation is enhanced by androgens. LNCaP cells cultured in androgen free media produced basal levels of FGF8b. Upon stimulation with androgens for 36 h, FGF8 levels in both the cell lysate and conditioned media were seen to increase at doses of 1 and 10 nM mibolerone (Figure 3a,b). Cells cultured with 10 nM mibolerone secreted significantly higher levels of FGF8b into the conditioned
media (P50.05) (Figure 3b). Over a period of 36 h, androgen stimulated cells (at a dose of 10 nM mibolerone) secreted increasing amounts of FGF8b (Figure 3c) as compared to cells cultured in DCC only (P50.005). These data suggest that the liganded AR is capable of modulating FGF8b expression. FGF8 gene promoter activity is upregulated by androgens acting via the AR A 1613 bp fragment corresponding to the upstream untranslated region of the FGF8 gene was subcloned into a luciferase reporter system (FGF8.luc) (Figure 4a). Sequence analysis confirmed that the fragment matched the published FGF8 5’ UTR (Acc:AF079532) and contained the first 169 bp of exon 1A of the published FGF8 coding sequence (Acc:AH006649) immediately upstream of the ATG initiation codon. Initial studies confirmed that the fragment contained promoter activity when compared to basic luciferase vector only (data not shown). Gemel et al., 1999 have also previously reported promoter activity in a similar 2 kb 5’ region of the murine FGF8 gene. We first transfected FGF8.luc into the natively AR expressing LNCaP and SC3 cell lines. In the presence of
Figure 2 FGF8b expression in CWR22 tumours is modulated by androgens. Tumour bearing mice were subjected to testosterone pellet implantation, castration or no treatment as in Materials and methods. At 4 weeks representative tumours from each group were harvested and serial sections stained for FGF8b, AR and PSA and counter-stained with haematoxylin. Sections with no primary antibody were used as negative controls (not shown). All images shown are magnified at 6200 Oncogene
Regulation of FGF8 expression in the prostate VJ Gnanapragasam et al
5072
Figure 3 Androgens regulate expression of FGF8b in LNCaP cells. (a) LNCaP cells cultured in the absence or presence of androgens (at doses of 1 and 10 nM Mibolerone) for 36 h at which point cell lysate was harvested and assayed for FGF8b. Loading was checked by subsequent blotting for a tubulin. (b) LNCaP cells cultured as previously described, the corresponding conditioned media were collected and the amount of secreted FGF8b was assayed. (c) LNCaP cells were exposed to androgens (10 nM mibolerone) for varying time intervals as shown. Conditioned media were then collected and assayed for FGF8b. Loading for each set of experiments in b and c was determined after assaying total protein content as described in Materials and methods. In all three figures basal levels of FGF8b expressed by cells maintained in DCC were given an arbitrary value of 1. All other results are expressed as fold induction over basal levels. Results shown are the mean and s.d. of three separate experiments
androgens, FGF8.luc activity was induced 2.5-fold in both cell lines compared to transfected un-induced cells (Figure 4b). To confirm that the AR was required for this induction, AR negative DU145 cells were transfected with FGF8.luc or PSA.luc with or without wild type AR (Figure 4c). In this experiment we observed that maximum activity of FGF8.luc required the presence of both the AR and exogenous androgens (2.5-fold). FGF8.luc was not induced by androgens in the absence of the AR. Similar studies performed using PSA.luc, demonstrated predictable induction of its promoter activity when the AR was co-transfected in the presence of androgens. From these results we hypothesise that AR modulation of FGF8 expression occurs at the gene promoter level. Anti-androgens inhibit AR mediated transactivation of the FGF8 promoter Clinical use of androgen ablation delays prostate cancer progression and induces a fall in serum PSA levels. We asked the question if anti-androgens were able to inhibit AR mediated FGF8.luc induction. DU145 cells cultured in DCC media were coOncogene
transfected with the AR and either FGF8.luc or PSA.luc and then cultured with androgens with or without bicalutamide at different doses (Figure 5a). Bicalutamide vehicle, ethanol, had no effect on AR induced FGF8.luc activity and bicalutamide itself had no effect on inherent FGF8.luc promoter activity. Bicalutamide was however able to significantly inhibit androgen induced FGF8.luc transcription. This inhibitory effect was comparable at 10- or 100-fold molar excess to androgen concentrations. A similar set of results was observed when PSA.luc was tested as a control with bicalutamide completely blocking AR mediated PSA promoter induction (Figure 5a). Bicalutamide was also able to block over-expression of endogenous FGF8b protein in LNCaP cells treated with androgens. Only basal levels of FGF8b were observed when these cells were exposed to androgens and 10-fold molar excess of bicalutamide (Figure 5b). Inherent activity of the FGF8 gene promoter is separate from the androgen response region The FGF8 promoter sequence has not previously been reported to contain androgen response elements
Regulation of FGF8 expression in the prostate VJ Gnanapragasam et al
5073
Figure 4 FGF8 gene promoter activity is modulated by androgens acting through the AR. (a) Schematic diagram of the proximal FGF8 gene promoter (FGF8.luc). (b) FGF8.luc was transfected into native AR expressing LNCaP and SC3 cells and then treated with androgens (M-mibolerone 10 nM). (c) FGF8.luc was transfected into AR negative DU145 cells with or without co-transfection of the AR. A similar experiment was performed with the PSA promoter (PSA.luc). Results are shown as fold induction above basal levels of the promoter cultured in DCC media and represent the means and s.d. of three independent experiments, each performed in triplicate
Figure 5 AR mediated transactivation of the FGF8 promoter is blocked by the anti-androgen bicalutamide. (a) DU145 cells were co-transfected with the AR and FGF8.luc and treated with androgens (10 nM mibolerone). The anti-androgen bicalutamide at 10 (100 nM) and 100 (1 mM) fold molar excess or the bicalutamide vehicle only (ethanol) was then added to the culture medium. A similar experiment using the PSA.luc reporter was performed. Results are shown as fold inductions above basal levels of the individual promoter cultured in DCC media and represent the mean and s.d. of three independent experiments, each performed in triplicate. (b) Western blot of FGF8b protein from the conditioned media of LNCaP cells treated with androgens (10 nM mibolerone) and in the presence of bicalutamide (100 nM) or the bicalutamide vehicle only. Basal levels of FGF8b expressed by cells in DCC were given an arbitrary value of 1. All other results are expressed as fold induction over this basal level Oncogene
Regulation of FGF8 expression in the prostate VJ Gnanapragasam et al
5074
(ARE). We therefore set out to define the region(s) within this fragment that was responsive to the AR. To
achieve this we designed deletion fragments and tested these for both promoter activity and androgen
Figure 6 The FGF8 promoter has separate activity and androgen responsive regions. (a) Deletion constructs derived from the 1.6 kb FGF8.luc fragment were co-transfected with the AR into DU145 cells in the presence and absence of androgens (M-mibolerone 10 nM). Activity was compared to pGL3 basic vector (negative control) and FGF8.luc (positive control). Results shown are fold inductions above pGL3 basic vector levels cultured in DCC media and represent the mean and s.d. of three independent experiments, each performed in triplicate. (b) F8luc1613-996 derived from the upstream region of FGF8.luc and cloned into the pGL3 promoter vector was co-transfected into DU145 cells together with the AR in the presence or absence of androgens (M-mibolerone 10 nM). Activity was compared to pGL3 promoter vector (negative control) and FGF8.luc (positive control). Results shown are fold inductions above pGL3 promoter vector levels cultured in DCC media and represent the mean and s.d. of three independent experiments, each performed in triplicate. (c) DU145 cells were co-transfected with the AR and F8luc1613-996, FGF8.luc or PSA.luc and treated with androgens (10 nM mibolerone). The anti-androgen bicalutamide at 10- (100 nM) and 100- (1 mM) fold molar excess was then added to the culture medium. In this experiment, results shown are fold inductions above each individual promoter cultured in DCC media and represent the mean and s.d. of three independent experiments, each performed in triplicate Oncogene
Regulation of FGF8 expression in the prostate VJ Gnanapragasam et al
5075
sensitivity (Figure 6a). Promoter activity remained intact in fragments 7996 and 7616 bp upstream of the ATG (F8luc616) but was lost in a fragment 7378 bp upstream of the ATG (F8luc378) suggesting that the basal promoter activity lay between nucleotides 7616 and 7378. All three deletion fragments tested however did not respond to androgen stimulation when compared to the original FGF8.luc fragment. We then cloned the region from 71613 to 7996 bp into the pGL3 promoter vector. Introduction of this fragment into the promoter vector resulted in the construct F8luc1613-996, which was confirmed to be androgen inducible (Figure 6b). Androgen was able to induce its activity by up to twofold above un-induced transfected cells. Furthermore, bicalutamide completely blocked androgen mediated induction of F8luc1613-996, similar to its effects on the FGF8.luc and PSA.luc promoters (Figure 6c). The androgen response region (F8luc1613-996) binds to the AR and contains putative androgen response elements We next asked if this androgen response region interacted in vivo with the AR. Nucleotide analysis of the entire region (spanning 617 bp) revealed three possible ARE sequences: 5’-GGGCCTggcTGTGCT-3’ (71559/71544), 5’-GCTACTgagAGTTTT (71111/ 71097) and 5’-AGGACCcacTGTCCC-3’ (71069/ 71055 in the reverse complementary strand) (Figure 7a). All three sequences identified were found to share at least 50% homology with the consensus DNA binding site for the AR (Figure 7a). To confirm in vivo interaction between AR protein and the identified androgen responsive region, we performed PCR on chromatin DNA specifically complexed to the AR protein. Three primer sets (F1R1, F2R2, F3R3) were designed to cover the FGF8 promoter androgen response region and produced amplicons of 150, 220 and 230 bp in size respectively (Figure 7b). PCR was then performed on AR immunocomplexed DNA obtained from LNCaP cells. Signal strength in noncomplexed chromatin was used to confirm that equal amounts of DNA had been loaded. With primer sets F2R2 and F3R3, there was no difference in the amplicon signal using DNA from LNCaP cells cultured in the presence or absence of androgens (Figure 7c). PCR with primer set F1R1, covering nucleotides 71593 to 71484, produced a much stronger signal when DNA from androgen treated LNCaP cells were used as a template (Figure 7c). This suggests the presence of a sequence within this fragment that interacts in vivo with the AR protein in a ligand dependent fashion. This fragment includes the sequence 5’-GGGCCTggcTGTGCT-3’ (71559/71544) described above, and its functional significance as a novel ARE requires further evaluation. Discussion The first aim of our study was to determine if FGF8 protein levels in prostate cancer could be modulated by
androgens. We used a commercially available antibody to determine levels of expression of human FGF8b. Immunohistochemistry demonstrated that FGF8b and AR were co-expressed in untreated prostate cancer as has been previously reported (Wang et al., 1999). In hormone refractory cancer however, FGF8b was expressed preferentially in AR positive cases. The CWR22 prostate xenograft is known to be exquisitely androgen sensitive and castration induces rapid cessation of proliferation (Myers et al., 1999). In this model we observed that castration reduced AR, PSA and FGF8b levels. Correspondingly, testosterone treatment significantly increased PSA and FGF8b expression. FGF8 is known to induce proliferation, most likely by initiating and propagating the cell cycle (Ghosh et al., 1996; Song et al., 2000). There is no current evidence however that FGF8 levels are altered during different phases of the cell cycle. We believe therefore that the changes observed in FGF8b expression is due to changes in the hormone status of CWR22 tumours rather than a change in the tumour proliferation rate. This observation was further supported by FGF8b induction studies in LNCaP prostate cancer cells known to express FGF8b mRNA (Ghosh et al., 1996). The peptide however contains signal sequences and most of the protein is secreted. We therefore assayed FGF8b levels in both the cell lysate and in the conditioned media. Similar to previous studies, we found basal levels of FGF8b expression in untreated LNCaP cells (Schmitt et al., 1996). Androgen stimulation however significantly induced FGF8b protein levels. It is interesting to note that using conventional RT – PCR, Schmitt et al. (1996) and Ghosh et al. (1996) failed to detect any changes in FGF8 mRNA expression in LNCaP cells following treatment with androgens. Taken together, our in vitro and in vivo data supports the model of androgen receptor mediated FGF8 expression that has been observed in murine mammary and human oesophageal cancer cells (Tanaka et al., 1992, 2001). FGF8 was first identified as an androgen induced growth factor but the mechanism of this induction is unconfirmed. Kapoun and Shackleford (1997) have reported that the FGF8 gene is selectively activated in mouse mammary tumour virus (MMTV) infected Wnt1 transgenic mice. Valve et al. (1998) have further observed that in the S115 mouse mammary tumour cell line, MMTV sequences containing steroid-regulated long terminal repeat (LTR) promoters were localized to DNA fragments adjacent to the FGF8 gene. In contrast, Kuriki et al. (2000) studying SC3 cells, found no retroviral insertions at the common integration sites adjacent to the FGF8 locus. Furthermore, androgen stimulation of AR positive KSE 1 cells (human oesophageal cancer) induced the expression of two FGF8 transcripts, namely FGF8b and FGF8f (Tanaka et al., 2001). MMTV pro-viral inserts are not known to occur in this cell line nor indeed in prostate cancer cell lines in which FGF8 has been detected. Promoter interaction with the AR was investigated by cloning a 1.6 kb fragment from the immediate 5’ Oncogene
Regulation of FGF8 expression in the prostate VJ Gnanapragasam et al
5076
Figure 7 The androgen response region (F8luc1613-996) binds to the AR and contains putative AREs. (a) Nucleotide sequence of the F8luc1613-996 fragment. Three putative ARE sites were identified and are underlined in this sequence. These sequences are shown aligned in comparison to the consensus AR binding sequence. The primer sequences used for PCR are also shown in this diagram (F – Forward primer, R – Reverse primer). (b) Three sets of primers were used to cover the FGF8 promoter androgen response region. These produced amplicons of 150 bp (F1R1), 220 bp (F2R2) and 230 bp (F3R3) using the FGF8.luc construct as DNA template. (c) Using these primers PCR was performed on chromatin DNA extracted from LNCaP cells cultured in the absence or presence of androgens (10 nM Mibolerone). 1 – Non-complexed DNA. 2 – Chromatin DNA immunocomplexed with AR antibody. Water was used as a negative control for each primer set
UTR region of the FGF8 gene into a reporter vector. Both endogenous and transfected AR induced promoter activity by up to 2.5-fold. We also observed that maximal activation of the promoter required the presence of both the AR and androgens and that this induction was abolished in the presence of antiandrogens. We did find however that in hormone refractory DU145 prostate cancer cells, transfection of the AR on its own had a mild inductive effect on FGF8.luc activity. This induction may have implications in hormone refractory disease as the AR is Oncogene
known to be activated even though the cells are insensitive to androgens. In this context, the AR would continue to induce FGF8 expression and drive tumourigenesis. Analysis of the 1.6 kb promoter fragment using deletion constructs revealed that fragments between 7616 and 7378 bp upstream of the ATG retained promoter activity while the androgen response region was located at least 996 bp upstream of the translation start codon. We were able to demonstrate recruitment of the AR to this region using ChIP assays. Specifically, sequence between
Regulation of FGF8 expression in the prostate VJ Gnanapragasam et al
5077
nucleotides 71593 and 71484 was implicated to either interact directly with or form a complex involving the AR in a ligand dependent manner. A putative novel ARE within this sequence has been identified (71559/ 71544) with homology to the optimal GGTACAnnnTGTTCTAR binding palindrome and to other known high affinity AREs (Roche et al., 1992; Claessens et al., 2001). Two other putative AREs (71111/71097 and 71069/71055) downstream of this sequence were also identified. Their functional significance is currently unclear as PCR covering this sequence failed to show enhanced signals in androgen treated AR immunocomplexed DNA. It is possible that these other sites are non-functional. Alternatively, they may require tandem binding of the first ARE for functional activation. Future work will be aimed at studying the interaction of the AR with the putative AREs described above in electromobility shift assays and mutagenesis assays and to determine if there are enhancer or repressor elements in other regions of the gene. In addition to specific palindromes, the DNA sequences flanking AREs are known to be essential for AR mediated transcriptional activity (Riegman et al., 1991; Murtha et al., 1993). An initial search for transcription factor binding sites failed to reveal likely candidates within the 71613/7996 upstream fragment. In the active promoter fragment (7616 bp upstream of the ATG), potential binding sites for Freac-4, EGR1, N-Myc as well as for TAF(II)135 and TAF(II)100 were identified using the TRANSFAC database (www.gene-regulation.com). Based on these findings it is reasonable to postulate that the AR in the presence of its ligand translocates to the nucleus where it recognizes and binds directly or interacts as part of a transcriptional complex with the FGF8 promoter. Interaction between the AR-ARE-transcription initiation complex may well be facilitated by the recruitment of AR co-activators. These co-activators can act as bridging molecules between steroid hormone receptors and general transcription factors resulting in increased activity of RNA polymerase II (Smith et al., 1996; Horwitz et al., 1996; Glass et al., 1997). Many coactivators have been shown to enhance AR transcriptional activity in prostate cancer cells (Yeh and Chang, 1996; Brady et al., 1999; Gnanapragasam et al., 2001). It is possible that such co-activators may be involved in ARE recognition and/or facilitate AR binding to nonclassical ARE sequences. In the present study we have reported androgen induction of the FGF8b isoform only. The mechanisms governing differential expression of FGF8 isoforms in the prostate is poorly understood. In prostate cancer cell lines, the b isoform is the predominantly expressed species (Ghosh et al., 1996) and has been consistently shown to be strongly tumourigenic (Rudra-Ganguly et al., 1998; Song et al., 2000). Recently, Valve et al., 2001 have demonstrated that the a and e isoforms of FGF8 are also expressed at high frequency in prostate cancer. Further work will be required to determine if the other FGF8 isoforms are equally induced by androgens and
to characterize the utilization of FGF8 isoform expression in prostate cancer development and progression. In conclusion we present evidence of androgen regulation of FGF8 expression in prostate cancer. This regulation requires the presence of the AR and acts at the level of the gene promoter and analysis of this promoter has suggested the presence of a candidate ARE sequence. We propose that AR dependent induction of FGF8, a potent mitogen, may be a factor in prostate cancer progression.
Materials and methods Patient samples Twenty patients with newly diagnosed prostate cancer as well as 15 with documented hormone relapsed cancer were identified from a pathology department database. Newly diagnosed cases were identified as those who had not had prior hormone therapy and were diagnosed at the time of surgery. Hormone relapsed cases were defined as patients with documented resurgence of serum PSA following androgen ablation therapy. All patients had presented with bladder outflow obstruction and required transurethral resection of the prostate. Following formalin fixation and embedding in paraffin, 5 mm sections were mounted on APES coated slides for further study. Sections of breast cancer and normal bronchus were identified from archival material. CWR22 prostate xenograft CWR22 prostate xenograft cells were received with thanks from Professor TG Pretlow (Case Western Reserve University Cleveland, Ohio, USA). Cell suspensions were mixed with Matrigel (Collaborative Research, USA) to give a total volume of 0.5 ml per injection (mean number of cells 56105). Single subcutaneous (s.c.) injections were given to 12 – 16 week old CD 1 male nude mice (Charles Rivers, UK) that had been housed at least 4 weeks prior to commencement of the study. Following implantation, tumours were allowed to grow for 6 weeks. At this point mice were randomly allocated into three groups. Under general anaesthesia, the first group (initial tumour volume 610.2+96.3 mm3) received s.c. implantation of 12.5 mg sustained released testosterone pellets (Innovative Research of America, Sarasota, Florida, USA) as has been previously described in this model (Wainstein et al., 1994). Pellets were located on the sub-scapular side opposite to the site of tumour injection. The second group (initial tumour volume 604.6+96.3 mm3) was castrated by bilateral orchidectomy and the third group (initial tumour size 560.6+117.6 mm3) were sham anaesthetized only. Initial tumour volumes between the groups were not significantly different (P40.05). The same person, using a standard technique, measured tumour volume at the time of treatment and then at weekly intervals for 4 weeks. Tumours were harvested at week 4 within each cohort. At harvest tumours were carefully dissected, denuded of host mouse tissue and immersed immediately in formalin for 48 h and paraffin embedded. 5 mm sections were cut from paraffin blocks and fixed on to APES coated slides for further study. All animal work was performed in a dedicated animal house according to Home Office guidelines. The project was performed as part of an existing Home Office licence. Oncogene
Regulation of FGF8 expression in the prostate VJ Gnanapragasam et al
5078
Antibodies
Cell culture and transient transfection
Mouse monoclonal FGF8b (500 mg/ml) was obtained from R&D systems (Abingdon UK), which specifically recognises FGF8b and c isoforms, and the human FGF8b isoform. We confirmed specificity of the antibody by probing for recombinant FGF8b (rhFGF8b) using Western blot analysis. Mouse monoclonal AR antibody (500 mg/ml) was obtained from Pharmingen (USA) and mouse monoclonal PSA antibody (500 mg/ml) from Novocastra (Newcastle-uponTyne, UK).
To investigate in vitro androgen induction of FGF8, LNCaP cells were pre-cultured in RPMI 1640 media (Life Technologies, USA) supplemented with 10% foetal calf serum (Full media-FM). On day 1, the media was removed and the cells washed with PBS. Media supplemented with dextran charcoal steroid depleted serum (DCC) was then added and the cells incubated for 24 h. On day 2, cells were washed and the media replaced with DCC, DCC M (DCC supplemented with 1 nM/10 nM mibolerone) or full media (FM) and incubated for a further 36 h. To investigate time dependent induction of FGF8b, LNCaP cells were starved for 24 h in DCC and then incubated with DCC, full media or DCC+10 nM mibolerone (DCC M) for 12, 24 or 36 h, as indicated. Transient transfections were performed using SuperfectTM reagent according to the manufacturer’s recommendation (Promega). In transient transfections of LNCaP and SC3 cells, 12-well plates were seeded with 80 000 LNCaP or 50 000 SC3 cells for 48 h before experiments. 250 ng of FGF8.luc and 200 ng of pCMV.bgal were transfected and cells incubated for 36 h with or without mibolerone. Cells were recovered and assayed for luciferase activity according to manufacturer’s recommendations (Promega). For transient transfections of DU145 cells, 24-well plates were seeded with 25 000 cells/well for 24 h prior to experiments. 50 ng of pCDNA3.AR or pCDNA3 only, 125 ng of FGF8.luc or PSA.luc and 200 ng of pCMV.bgal were co-transfected. As before, cells were incubated for 36 h with or without mibolerone before assaying luciferase and b galactosidase activities. Individual luciferase activities were corrected by the corresponding b galactosidase activities to obtain the relative activity. b galactosidase assays were typically performed in a 96-well plate (Corning) and A420 values were obtained using the MR5000 plate reader (Dynatech). For studies involving antiandrogens, bicalutamide (Astra-Zeneca, UK) at a concentration of 50 mM in ethanol was diluted in DCC M media to working solutions of 100 nM and 1 mM.
Plasmid and vectors A 10 kb genomic clone containing the human FGF8 gene was a kind gift from Dr Clive Dickson (ICRF, UK). From this larger clone restriction mapping revealed a 1613 bp fragment corresponding to the upstream un-translated region of the FGF8 gene. Sequence analysis of this fragment revealed GC rich regions without evidence of a TATA box. This 1.6 kb fragment was then subcloned into the BglII/NcoI site of the pGL3-basic luciferase reporter vector (Promega), hereafter referred to as FGF8.luc, for subsequent in vitro transfection experiments. A PSA reporter construct containing two androgen response elements (PSA.luc), human AR cloned into the pCDNA3 vector (pCDNA3.AR) and pCMV.bgal have been previously described (Brady et al., 1999). Empty pSG5 vector was used to normalize the amount of DNA transfected in each experiment. For studies investigating the androgen responsive region of the FGF8 promoter, the fragment of interest was cloned into the pGL3-promoter vector (Promega). Immunohistochemistry Paraffin sections were baked overnight at 508C, deparaffinized in xylene and rehydrated through graded alcohols. Endogenous peroxidase activity was blocked by incubating in methanol peroxide (30% H2O2) for 10 min. Antigen retrieval by microwave treatment was performed using a hot start variation in which sections were placed in 1.5 l of preheated (958C) 0.01 M sodium citrate buffer (pH 6) and then re-heated for 6 min. Following this, non-specific binding was blocked by incubating the slides in 10% rabbit serum (DAKO) for 30 min followed by incubation with the relevant primary antibody overnight. The next day, bound antibody was detected with biotinylated antimouse immunoglobulin and visualized with 3,3’-diaminobenzidine tetrahydrochloride (Sigma). Breast cancer and bronchus sections were used as positive and negative controls for FGF8b expression respectively. Sections incubated without any primary antibody were also employed as negative controls. All sections were counterstained with haematoxylin. Sample sections were first previewed by two independent observers and inter-observer agreement obtained regarding a grading system. This template was then used for all cases studied. In clinical cancers, the level of FGF8b and AR expression was assumed to correlate with the immunoreactivity signal generated as either absent (7) or present (+). In the CWR22 model, expression of the AR, PSA and FGF8b was assumed to correlate directly with the strength of the immunoreactivity signal and graded as absent (7), positive (+), or strongly positive (++). For this analysis at least three separate areas within each slide were examined and positivity defined as the presence of a signal in at least two out of three areas. Oncogene
Analysis of FGF8b protein To asssay FGF8b levels in LNCaP cell lysate, androgen treated LNCaP cells were harvested with SDS loading buffer (0.125 M Tris pH 6.8, 2% sodium dodecylsulphate, 10% glycerol, 10% b mercaptoethanol and 0.01% bromophenol blue). To assay secreted FGF8b, media conditioned by androgen treated LNCaP cells was harvested and the protein content measured by the Bradford assay method (BCA, Pierce). Aliquots of conditioned media were then centrifuged to remove cell debris and heparin-binding proteins isolated from the supernatant by overnight incubation with 100 ml of heparin-sepharose (1 : 1 slurry, Pharmacia). Heparin-sepharose beads were then precipitated and washed three times with 20 mM Tris-HCL, pH 7.4, 300 mM NaCl. Bound proteins were released by heating to 1008C for 5 min in SDS sample buffer (0.125 M Tris pH 6.8, 2% sodium dodecylsulphate, 10% glycerol, 10% b mercaptoethanol and 0.01% bromophenol blue). 20 ml aliquots of sample in SDS leading buffer were subjected to 10% SDS – PAGE before protein transfer to nitro-cellulose membranes (Hybond C, Amersham UK). Filters were first blocked with 5% dried milk at room temperature for 1 h before incubation with mouse monoclonal FGF8b antibody (1 mg/ml) in sealed bags at room temperature for 1 h. Filters were then incubated with horseradish peroxidase conjugated rabbit anti-mouse secondary antibody (R&D Systems, Abingdon, UK) before visualization with ECL reagents (Amersham, UK). For
Regulation of FGF8 expression in the prostate VJ Gnanapragasam et al
5079
induction analysis, Western signals were analysed and quantified using a digital scanner (Molecular Dynamics, Chesham, Bucks, UK). Densitometry readings were taken and the values expressed as fold induction when compared to cells maintained in DCC (control). As a negative control nonconditioned media was used and probed for FGF8b expression exhibiting no signals. Deletion constructs FGF8.luc was subjected to a combination of restriction endonuclease digestion and exonuclease digestion to obtain the promoter constructs reported. F8.luc1613-996 was derived from digesting FGF8.luc with BglII and HindIII, sub cloning the released fragment into pBluescript (Stratagene) and then released again with XhoI-XhoI before sub cloning into the pGL3 promoter vector. F8.luc996, F8.luc616 and F8.luc378 were obtained by digestion of KpnI-XhoI linearised FGF8.luc using Exonuclease III (Promega) according to the manufacturer’s protocol. Samples were digested at 308C for 1 min intervals at which point aliquots were removed and the reaction stopped in S1 nuclease mix (50 IU S1 nuclease and 15 ml of S1 106 buffer (Promega)). Following Klenow treatment, digested constructs were ligated and transformed. Selected colonies were screened for inserts and successful clones midi-prepped to obtain working stocks of DNA for transfection experiments. All derived fragments were checked by sequence analysis. Chromatin immunoprecipitation (ChIP) and PCR analysis LNCaP cells were grown on 150-mm dishes (Corning) in full media for 2 days until approximately 56106 cells were present. Cells were then transferred to DCC medium for 16 h prior to steroid treatment. After 16 h, the media was replaced with DCC media supplemented with or without androgens (10 nM mibolerone). Following treatment, LNCaP cells were treated with formaldehyde, added directly to culture medium (to a final concentration of 1%), at room temperature for 10 min, to cross-link histone proteins to DNA. Soluble chromatin was made as follows: cells were washed and detached from dish by scraping following addition of ice-cold phosphate-buffered saline supplemented with 25 mg ml71 leupeptin, 25 mg ml71 aprotinin and 25 mg ml71 pepstatin, and pelleted by centrifugation for 4 min at 700 g. The resultant cell pellet was then lysed in lysis buffer (50 mM Tris-HCl, pH 8.1; 1% SDS, 10 mM EDTA, 1 mM PMSF, 0.8 mg ml71 pepstatin, 0.6 mg ml71 leupeptin and 0.6 mg ml71 aprotinin) and sonicated. Samples were then centrifuged at 13 000 r.p.m. for 10 min and supernatant decanted and diluted 10-fold in dilution buffer (25 mM Tris, pH 8.1; 140 mM NaCl, 1% SDS, 3 mM EDTA, 1 mM PMSF, 0.8 mg ml71 pepstatin, 0.6 mg ml71 leupeptin and 0.6 mg ml71 aprotinin). To pre-clear chromatin solution, 60 ml salmon sperm DNA/protein A agarose beads (Upstate Biotechnology) was added to each sample and agitated for 30 min at 48C. Beads were pelleted by brief centrifugation and supernatant collected. For immunoprecipitation, 2 mg of
AR antibody was added to 1 ml of the purified chromatin sample and incubated overnight at 48C. Immunocomplexes were recovered by adding 60 ml salmon sperm/protein A agarose for 1 h at 48C with agitation. Beads were washed sequentially for 5 min each in 10 ml TSE (0.1% SDS, 1% Triton X-100, 2 mM EDTA, 20 mM Tris-HCl, pH 8.1) plus 150 mM NaCl, TSE plus 500 mM NaCl, and buffer III (0.25 M LiAc, 1% Nonidet P-40, 1% deoxycholate, 1 mM EDTA, 10 mM Tris-HCl, pH 8.1) and then washed three times with TE (pH 8). Immunocomplexes were eluted by adding 250 ml elution buffer (1% SDS and 0.1 M NaHCO3) to beads and subsequently heated for 4 h at 648C to reverse formaldehyde-induced cross-links. AR immunocomplexed DNA was then recovered by phenol/chloroform extraction, ethanol precipitation and resuspended in 50 ml TE. PCR Primer sets were designed to overlap and span the androgen response region of the FGF8 promoter and give bands of approximately 150 – 230 bp. Primer set 1: Forward: 5’-AGTTGGAAAGATGGGGCACA-3’, Reverse: 5’-GTCTTCACTTACAACCTCCC-3’. Primer set 2: Forward: 5’-ATGGAGGGTCACTGGGGA-3’, Reverse: 5’-TTCTAGGCAGGGGCTGGT-3’. Primer set 3: Forward: 5’-CCTTCTCCTATGCCCCAAGC-3’, Reverse: 5’-AAGCTTGAGGAGCACCCCGA-3’. These primers were first evaluated using the FGF8.luc construct as DNA template. Quantitative PCR was then performed with 10 ml of eluted AR immunocomplexed DNA, BioTaq DNA polymerase (Bioline) and [a32P]dATP (ICN, Oxon, UK). PCR was also performed on unprecipitated chromatin as a positive control and to correct for input volume. Amplification was carried out for 35 cycles (25 cycles for unprecipitated chromatin input lanes) with denaturation at 948C for 1 min, annealing at 648C for 30 s and extension at 728C for 1 min. PCR products were resolved, dried and then exposed to X-ray film for 2 – 12 h. A water control was also used in each experiment.
Abbreviations AR, androgen receptor; ARE, androgen response element; FGF8, fibroblast growth factor 8; ChIP, chromatin immunoprecipitation; CWR22, Case Western Reserve Prostate Xenograft strain 22; MMTV, mouse mammary tumour virus; PCR, polymerase chain reaction; PSA, prostate specific antigen; SHR, steroid hormone receptor
Acknowledgments This project was funded by the Cancer Research Campaign of the UK, grant no SP2503/0101. We thank Dr Clive Dickson (Imperial Cancer Research Fund, London, UK) for the human FGF8 genomic clone, Professor TG Pretlow (Case Western Reserve University, Cleveland, Ohio, USA) for the CWR22 cells, Professor DR Newell (Cancer Research Unit, University of Newcastle-upon-Tyne) for assistance with the animal studies. We thank Susan Cook and Louise McCarthy for expert technical assistance.
References Bower M and Waxman J. (1997). Sem. Cancer Biol., 8, 3 – 9. Brady ME, Ozanne DM, Gaughan L, Waite I, Cook S, Neal DE and Robson CN. (1999). J. Biol. Chem., 274, 17599 – 17604.
Claessens F, Verrijdt G, Schoenmaker E, Haelens A, Peeters B, Verhoven G and Rombauts W. (2001). J. Steroid Biochem. Mol. Biol., 76, 23 – 30.
Oncogene
Regulation of FGF8 expression in the prostate VJ Gnanapragasam et al
5080
Crossley PH, Minowada G, MacArthur CA and Martin GR. (1996). Cell, 84, 127 – 136. Daphna-Iken D, Shankar DB, Lawshwe A, Ornitz DM, Shackleford GM and MacArthur CA. (1998). Oncogene, 17, 2711 – 2717. De Vere White RW, Meyers F, Chi SG, Chamberlain S, Siders D, Lee F, Stewart S and Gumerlock PH. (1997). Eur. Urol., 31, 1 – 6. Dorkin TJ, Robinson MR, Marsh C, Bjartell A, Neal DE and Leung HY. (1999). Oncogene, 18, 2755 – 2761. Gemel J, Gorry M, Ehrlich GD and MacArthur CA. (1996). Genomics, 35, 253 – 257. Gemel J, Jacobsen C and MacArthur C. (1999). J. Biol. Chem., 274, 6020 – 6026. Ghosh AK, Shankar DB, Shakleford GM, Wu K, T’Ang A, Miller GJ, Zheng J and Roy-Burman P. (1996). Cell Growth Diff., 7, 1425 – 1434. Glass CK, Rose DW and Rosenfeld MG. (1997). Curr. Opin. Cell. Biol., 9, 222 – 232. Gnanapragasam VJ, Leung HY, Pulimood AS, Neal DE and Robson CN. (2001). Br. J. Cancer, 85, 1928 – 1936. Greenlee RT, Murray T, Bolden S and Wingo PA. (2000). Can. J. Clin., 50, 7 – 33. Horwitz KB, Jackson TA, Bain DL, Richer JK, Takimoto GS and Tung LS. (1996). Mol. Endocrinology, 10, 1167 – 1177. Ittman M and Mansukhani A. (1997). J. Urol., 157, 351 – 356. Kapoun AM and Shackleford GM. (1997). Oncogene, 14, 2985 – 2989. Koivisto PA and Helin HJ. (1999). J. Pathol., 189, 219 – 223. Kouhara H, Koga M, Kasayama S, Tanaka A, Kishimoto T and Sato B. (1994). Oncogene, 9, 455 – 462. Kuriki K, Kamiakito T, Yoshida H, Saito K, Fukayama M and Tanaka A. (2000). Cell Mol. Biol., 46, 1147 – 1156. Lee SM, Danielian PS, Fritzsch B and McMahon AP. (1997). Development, 124, 959 – 969. Leung HY, Dickson C, Robson CN and Neal DE. (1996). Oncogene, 12, 1833 – 1835. MacArthur CA, Lawshe A, Shankar DB, Heikinheimo M and Shackleford GM. (1995). Cell Growth Diff., 6, 817 – 825. Merril RM, Potosky AL and Feuer EJ. (1996). J. Natl. Can. Inst., 88, 1683 – 1685. Myers RB, Oeschlager DK, Coan PN, Frost AR, Weiss HL, Manne U, Pretlow TG and Grizzle WE. (1999). J. Urol., 161, 945 – 949.
Oncogene
Murtha P, Tindall DJ and Young CY. (1993). Biochemistry, 32, 6459 – 6464. Riegman PH, Vlietstra RJ, van der Korput J, Brinkmann AO and Trapman J. (1991). Mol. Endocrinology, 5, 1921 – 1930. Roche PJ, Hoare S and Parker MG. (1992). Mol. Endocrinology, 6, 2229 – 2235. Rudra-Ganguly N, Zheng J, Hoang AT and Roy-Burman P. (1998). Oncogene, 16, 1487 – 1492. Schmitt FJ, Hearn MT and Risbridger GP. (1996). J. Steroid Biochem. Mol. Biol., 57, 173 – 178. Smith CL, Onate SA, Tsai MJ, McDonnel PD and O’Malley BW. (1996). Proc. Natl. Acad. Sci. USA, 93, 8884 – 8888. Song Z, Powell W, Kasahara N, Van Bokhoven A, Miller G and Roy-Burman P. (2000). Cancer Res., 60, 6730 – 6736. Story MT, Hopp KA, Molter M and Meier DA. (1994). Growth Factors, 10, 269 – 280. Tanaka A, Miyamoto K, Minamino N, Takeda M, Sato B, Matsuo H and Matsumoto K. (1992). Proc. Natl. Acad. Sci. USA, 89, 8928 – 8932. Tanaka A, Furuya A, Yamasaki M, Hanai N, Kuriki K, Kamiakito T, Kobayashi Y, Yoshida H, Koike M and Fukuyama M. (1998). Cancer Res., 58, 2053 – 2056. Tanaka S, Ueo H, Mafune K, Mori M, Wands JR and Sugimachi K. (2001). Dig. Dis. and Sci., 46, 1016 – 1021. Valve EM, Tasanen MJ, Ruohola JK and Harkonen PL. (1998). Biochem. and Biophys. Res. Comm., 250, 805 – 808. Valve EM, Nevalainen MT, Nurmi MJ, Laato MK, Martikainen PM and Harkonen PL. (2001). Lab. Invest., 81, 815 – 826. Van der Kwast T, Schalken J, Ruizveld de Winter J, van Vroonhoven C, Mulder E, Boersma W and Trapman J. (1991). Int. J. Cancer, 48, 189 – 193. Wainstein MA, He F, Robinson D, Kung HJ, Schwartz S, Giaconia JM, Edgehouse NL, Pretlow TG, Bodner DR and Kursh ED. (1994). Cancer Res., 54, 6049 – 6052. Wang Q, Stamp GW, Powell S, Abel P, Laniado M, Mahony C, Lalani EN and Waxman J. (1999). J. Clin. Pathol., 52, 29 – 34. Wu J, Payson RA, Lang JC and Chiu IM. (1997). J. Steroid Biochem. Mol. Biol., 62, 1 – 10. Yeh S and Chang C. (1996). Proc. Natl. Acad. Sci. USA, 93, 5517 – 5521. Yoshiura K, Leysens NJ, Chang J, Ward D, Murray JC and Muenke M. (1997). Am. J. Med. Gen., 72, 354 – 362.
