Oncogene (1998) 16, 2647 ± 2656  1998 Stockton Press All rights reserved 0950 ± 9232/98 $12.00 http://www.stockton-press.co.uk/onc FGF signaling activates STAT1 and p21 and inhibits the estrogen response and proliferation of MCF-7 cells

Mark R Johnson, Catherine Valentine, Claudio Basilico and Alka Mansukhani

Department of Microbiology and the Kaplan Cancer Center, New York University School of Medicine, New York, NY 10016, USA

Normal breast tissue as well as most breast tumors are until recently was that its mitogenic e€ect is partially dependent on estrogen for growth. Breast tumors often mediated by the induction of growth factor signaling progress to a hormone-independent state which is pathways (Dickson et al., 1986). Recent data however associated with poor prognosis. It has been proposed indicate that estrogen action may directly modulate the that activation of growth factor signaling pathways in activity of cell cycle regulatory genes and cause the tumor cells may free them from hormonal control. phosphorylation of Rb (Altucci et al., 1996). Certain growth factors can mimic estrogen responses by It has been known for some time that growth factors activating the estrogen receptor via its phosphorylation such as EGF and IGF, which are probably derived by mitogen-activated protein (MAP) kinase. In this from the surrounding stromal cells, can act on report, however, we show that ®broblast growth factor epithelial cells and modulate their growth. These (FGF), despite activating MAP kinase, is growth- growth factors are able to activate the estrogen inhibitory for estrogen-dependent MCF-7 breast cancer receptor and promote transcription from an ERE- cells. MCF-7 cells treated with FGFs exhibit slower containing promoter in the absence of estrogen. In the growth than controls in both the presence and absence of reproductive tract EGF can mimic the physiological estrogen, with a concomitant increase in the number of e€ects of estrogen (Ignar-Trowbridge et al., 1992; cells in G0/G1. Expression of a constitutively activated Curtis et al., 1996). This ligand-independent activation FGF receptor in these cells further decreases their of ER by IGF or EGF has been shown to be mediated growth rate, which is no longer in¯uenced by FGF through phosphorylation of a serine residue at position treatment. Activation of the FGF signaling pathway also 118 in ER by activated MAP kinase (Kato et al., 1995; reduces the induction of an estrogen-responsive CAT Bunone et al., 1996). reporter plasmid by estrogen, an e€ect which appears to A common feature associated with breast tumor be independent of serine 118 in the estrogen receptor, a progression is a loss of hormone-dependent growth of MAP kinase target site. The inhibitory e€ects of FGF the tumor cells. It has been suggested that the tumor are probably mediated through the sustained induction of cells, which are epithelial in origin, may acquire an the cyclin kinase inhibitor p21/WAF1/CIP1, which is autocrine or paracrine growth factor-stimulated path- upregulated at the mRNA and protein level by FGF. way and thus free themselves from hormonal control FGF treatment also results in the phosphorylation of (Dickson and Lippman, 1995; Chalbos et al., 1994). STAT1. This upregulation of p21 and phosphorylation of Signaling through a variety of growth factor receptors STAT1 is not detectable in T47D breast cancer cells including TGFa, EGF, IGF and FGF has been upon which FGF has no inhibitory e€ect. implicated in such autocrine growth (Ethier, 1995). The ®broblast growth factors (FGFs) are a family of Keywords: ®broblast growth factor; estrogen receptor; angiogenic factors that are potent mitogens for MAP kinase; STAT1; MCF-7 growth inhibition; breast ®broblast and endothelial cells (Basilico and Moscatel- cancer li, 1992), and have been implicated in epithelial-stromal interactions in hormone-responsive tumors. It has been well established that FGF signaling plays an important role in MMTV-induced mouse mammary tumors, since Introduction both FGFs and their receptors have been found to be activated in MMTV-induced mammary tumors (Mur- The principal mitogen for breast epithelial cells is the akami et al., 1990; Muller et al., 1990; Shackleford et al., steroid hormone estrogen but growth factors also play 1993; MacArthur et al., 1995). On the other hand, while an important role in controlling mammary gland some reports support the hypothesis that human breast development, morphogenesis, and breast cancer pro- tumor progression may be mediated by the autocrine gression (Dickson and Lippman, 1995; Chalbos et al., activation of the FGF signaling pathway (McLeskey et 1994). Estrogen acts by activating the estrogen receptor al., 1993; Adnane et al., 1991), more recent data point to (ER), a member of the nuclear receptor superfamily a negative regulatory role for these factors (Fenig et al., that dimerizes upon ligand binding, and regulates 1997; Wang et al., 1997). Furthermore, FGF-1 and transcription from ERE (estrogen response element)- FGF-2 levels are lower in breast tumor biopsies than in containing promoters in a cell-type speci®c manner normal breast, suggesting that in breast tissue these (Tsai and O'Malley, 1994). The mechanism by which factors may have an inhibitory role that is lost as the estrogen acts as a mitogen is still unclear; the consensus tumors progress (Bansal et al., 1995; Yiangou et al., 1997). Therefore we decided to investigate the cross-talk between FGF signalling and estrogen-response pathway Correspondence: C Basilico and A Mansukhani in breast epithelial cells. Received 13 August 1997; revised 18 December 1997; accepted 19 In this report we analyse the e€ect of FGF signaling December 1997 on the estrogen-responsive breast carcinoma cell lines Inhibitory effects of FGF on MCF-7 cells MR Johnson et al 2648 MCF-7 and T47D. We show here that FGF signaling although other ER-negative lines showed the presence is growth-inhibitory for MCF-7 cells, and that it can of FGF-1, FGF-2, FGF-5, or FGF-4 (Table 1). FGF-8 down-regulate the estrogen response in these cells was originally identi®ed from the SC-3 murine breast despite activating the MAP kinase pathway. Activa- carcinoma line and found to be androgen-inducible in tion of FGF signaling in MCF-7 cells results in a some human breast cancer cells (Payson et al., 1996). dramatic and sustained upregulation of the cyclin MCF-7 cells were positive for FGF-8 expression while dependent kinase inhibitor p21/WAF1 and an accu- T47D cells were negative. All the other estrogen- mulation of cells in G1. FGF treatment also induces independent cell lines also showed the presence of the phosphorylation of STAT1 (Signal Transducers FGF-8 mRNA. Primary ®broblasts derived from a and Activators of Transcription), a protein in the JAK- breast tumor biopsy were also analysed alongside the STAT signaling cascade that promotes the expression total tumor RNA for FGF expression. As expected, of a speci®c set of genes (Darnell et al., 1994; Ihle and the stromal ®broblasts appear to be the primary source Kerr, 1995). In contrast, the upregulation of p21 of FGF-1, FGF-2, FGF-5, FGF-7 and FGF-9 detected mRNA and protein, and the phosphorylation of in the tumor sample. A summary of the analysis is STAT1 is not seen in the T47D cells where FGF has shown in Table 1. no inhibitory e€ect. These results are consistent with The FGF receptors (FGFR) are encoded by four the hypothesis that FGF signaling antagonizes the distinct but homologous genes (FGFR-1, FGFR-2, estrogen response of MCF-7 breast cancer cells and FGFR-3 and FGFR-4). These receptors are mem- inhibits their proliferation by inducing growth-inhibi- brane-spanning tyrosine kinases of the immunoglobulin tory molecules. (Ig) superfamily with two or three Ig-loops in the extracellular region (Johnson and Williams, 1993). Splicing alternatives further give rise to receptor isoforms that can be identi®ed by RT ± PCR (Ittman Results and Mansukhani, 1997). The ligand-binding ability of these receptors is non-speci®c and overlapping for the Expression of FGFs and FGF receptors in breast cancer di€erent FGFs, with each receptor able to bind more cells than one growth factor (Ornitz et al., 1996). The only The FGF family of growth factors consists of nine strict speci®city is seen in a splice variant of FGFR-2 members described so far. We analysed the repertoire called FGFR-2(IIIb), which binds almost exclusively of FGFs expressed in several breast cell lines by RT ± FGF-7/keratinocyte growth factor. It is generally true PCR in order to rule out the involvement of autocrine that the two splice variants of FGFR-2 are expressed FGF e€ects in the MCF-7 and T47D cells. A set of in a mutually-exclusive cell type-speci®c manner, with primers speci®c for each FGF was designed that was FGFR-2(IIIb) in epithelial cells and FGFR-2(IIIc) in previously used to con®rm the array of FGFs present ®broblasts. However, several breast epithelial lines in prostate stromal and epithelial cells (Ittman and express both splice variants of receptor 2 (Luqmani Mansukhani, 1997). The PCR products were hybri- et al., 1996). MCF-7 and T47D cells express all four dized with DNA probes speci®c for each of the FGFs FGF receptors. Both splice variants of receptor 2 are in order to con®rm the presence or absence of a given expressed in MCF-7 cells. Somewhat surprisingly, all PCR product. There was no expression of FGFs 1 ± 7 other cell lines we examined express only the FGFR- or 9 in the two estrogen-responsive cell lines examined 2(IIIc) form.

Table 1 Expression of FGFs and FGF receptors in breast cells by RT ± PCR analysis FGF 123456789 MCF-7 7 7 7 7 7 7 7 + 7 T47D 7 7 7 7 7 7 7 7 7 HBL100 + + 7 7 n.d. 7 7 + 7 BT549 + + 7 + n.d. 7 7 + 7 MDAMB231 + + 7 7 ++ 7 7 + 7 H578TBst +++ 7 7 7 ++ 7 ++ + 7 Hs578T + ++ 7 + ++ 7 + ++ + Fibroblasts +++ ++ 7 7 ++ 7 +++ 7 ++ Tumor tissue + + 7 7 + 7 + 7 +

FGFR R1 (IIIb) R1 (IIIc) R2 (IIIb) R2 (IIIc) R3 (IIIb) R3 (IIIc) R4 MCF-7 7 + ++ ++ 7 + + T47D 7 7 7 ++ 7 + + HBL100 7 + 7 ++ 7 ++ + BT549 7 + 7 ++ 7 +/7 +/7 MDAMB231 7 + 7 ++ 7 +/7 + H578TBst 7 +/7 7 ++ +/7 + ++ Hs578T 7 + 7 ++ 7 +/7 + MCF-7 and T47D are estrogen-dependent tumor cell lines, while BT549, MDAMB231, and Hs578T are estrogen-independent tumor lines. HBL100 is an estrogen-independent line from mammary epithelial cells. H578TBst is the normal counterpart of Hs578T. The ®broblast sample was derived from a malignant breast biopsy sample (tumour tissue). (7), no band visible after overnight exposure; (+/7), band visible after overnight exposure; (+), band visible in 1 h exposure; (++) band visible on ethidium bromide stained gel; (+++) strong band on ethidium bromide stained gel; (n.d.), not done Inhibitory effects of FGF on MCF-7 cells MR Johnson et al 2649 of the cells had a G1/G0 DNA content while treatment FGF inhibits the growth of MCF-7 cells with 10 ng/ml FGF-2 increased this fraction to 73%. MCF-7 cells are dependent on estrogen for growth. The percentage of cells in S-phase was concomitantly Since fetal calf serum contains steroid hormones, the reduced from 27% to about 14%. The accumulation of cells were grown in phenol red-free medium containing cells with a G1 DNA content was somewhat lower in 5% charcoal-stripped serum to simulate estrogen- FGF-4 treated cells although the e€ect was still depleted conditions. Figure 1 shows that cells in this apparent (data not shown). The addition of estradiol medium grow slowly, and in the presence of 10 ng/ml was growth stimulatory with an increase in S phase FGF-4 there is a further reduction of growth. As and G2/M cells. Addition of estradiol and FGF-2 to expected, addition of 10 nM b-estradiol had a potent the cells showed that estrogen counteracted the growth stimulatory e€ect. However, the addition of inhibitory e€ect of FGF-2 (Table 2). 10 nM b-estradiol and FGF-4 together slowed the To ensure that the inhibitory e€ect we observed was growth of the cells compared to estrogen alone (Figure mediated by FGF signaling through the FGF receptor 1a). This e€ect was not seen in the estrogen- and not by a transmodulatory mechanism, we independent breast cell lines HBL100, H578T, and expressed a constitutively activated FGFR-2 in MCF- BT549 where the addition of FGF-4 or FGF-2 at 1 or 7 cells. The activated receptor has a cysteine to 10 ng/ml had no e€ect on the growth rate (data not arginine mutation at position 382 in the transmem- shown). In T47D cells, another estrogen-dependent cell brane domain. We have previously shown that this line which expresses approximately one-fourth as much receptor is capable of transforming NIH3T3 cells and estrogen receptor as the MCF-7 cells (Horwitz et al., is tyrosine-phosphorylated even in the absence of 1978), FGF-2 (data not shown) and FGF-4 had no ligand (Li et al., 1997). Several clones were isolated inhibitory e€ect and possibly a slight stimulatory e€ect and those expressing the highest levels of FGFR-2/ (Figure 1b). C382R were selected for further study. Clone 1021 #18 DNA content analysis showed that in untreated which expressed the highest level of the mutant cycling cells in estrogen deprived medium, about 63% receptor grew visibly slower than the parental cells and had a longer doubling time. The cells were still estrogen-responsive and the addition of FGF-4 had no further repressive e€ect on growth (Figure 1c). a The inhibitory e€ect of FGF on MCF-7 cells can also be seen at the level of DNA synthesis. Cells were kept in phenol red-free DMEM containing 5% charcoal-stripped serum for 48 h and treated with the indicated doses of FGF-4 and estradiol. Twelve hours later the cells were pulsed with 3H-thymidine and the radioactivity precipitated by TCA was counted. Figure 2 shows that in the absence of estrogen, FGF-4 treatment at 1 and 10 ng/ml reduces the incorporated counts two and fourfold respectively. This e€ect is still manifest in the presence of estrogen, but the inhibitory b e€ect is diminished, again indicating that estrogen and FGF have opposing e€ects on these cells. FGF treatment had no e€ect on DNA synthesis in 1021 #18 cells, which is consistent with the growth data (data not shown).

FGF can activate the MAP kinase pathway in MCF-7 cells Activation of erk1 and erk2 are known downstream c e€ects of FGF signaling in ®broblasts. We checked the kinetics of MAP kinase activation in our MCF-7 cells in response to FGF by an in vitro kinase assay. Cells were kept in 0.5% CS-FCS overnight (serum does not contain signi®cant levels of FGF activity). FGF-2 or

Table 2 DNA content analysis of MCF-7 cells in the presence of FGF or estrogen Treatment G0/G1 S G2/M Figure 1 Growth curves. Cells were grown for 4 days in phenol no treatment 63 27 9.1 red-free DMEM supplemented with 5% charcoal-stripped fetal FGF-2 10 ng/ml 73 14 12 calf serum, then plated at 104 cells per well in 24 well plates. E2 10 nM 46 33 20 Twenty-four hours later (day 0), cells were trypsinized and E2+FGF-2 55 29.2 15 counted. Medium and supplements were changed on the indicated counting days. E2=17b-estradiol. (a) MCF-7 cells, (b) T47D cells, MCF-7 cells were grown in phenol red-free medium containing 5% (c) MCF-7/1021 #18 cells. MCF-7 cells treated with estrogen were charcoal stripped serum for 2 days and then treated as described for shown for comparison 24 h. Cells were harvested and analysed for DNA content Inhibitory effects of FGF on MCF-7 cells MR Johnson et al 2650

a – E2 EGF FGF2 FGF4 Elk–1

b Active MAPK

Erk2

Time: 0 15'5h22h 5h 22h 15' 5h 22h E2: – ––– + ++++ FGF4: – +++–– ++ + c Figure 2 E€ect of FGF and estrogen on DNA synthesis in MCF-7 cells. MCF-7 cells were treated overnight with 0, 1, or Time 0 5' 30' 60'18h 30' 10 nM estradiol and 0, 1, or 10 ng/ml FGF4. The following day Treatment – FFFFI the cells were labeled for 6 h with 1 mCi/ml 3H-thymidine. Incorporated radioactivity was recovered by cold trichloroacetic Active MAPK acid precipitation and counted Figure 3 Activation of MAP Kinase by FGF in MCF-7 cells. (a) In vitro kinase assay. Cells were treated with 100 nM estradiol, 100 ng/ml FGF-2, 100 ng/ml FGF-4, or 100 ng/ml EGF for 30 min and lysed. Active MAP kinase was immunoprecipitated, FGF-4 at 100 ng/ml was applied for 30 min and cells and elk-1 was added as a substrate. Phosphorylated elk-1 was were harvested. To assess MAP kinase activity, active detected using an anti-phospho-elk-1 antibody. (b) Time course of MAP kinase activation by Western analysis of whole cell lysates. MAP kinase was immunoprecipitated and used in a MCF-7 cells were treated as indicated, lysed, and subjected to kinase assay with elk-1 as substrate. The kinase assay SDS ± PAGE. Active MAP kinase and MAP kinase (erk2) were was subjected to Western blotting with an antibody detected by immunoblotting. FGF-2 was at 10 ng/ml and that speci®cally recognizes phosphorylated elk-1. estrogen at 10 nM.(c) Time course of MAP kinase activation in T47D cells. Cells were treated with 10 ng/ml FGF-2 (F) for the Figure 3a shows that FGF-2 and FGF-4 can activate indicated times or IFN-g (I) as a control. Whole cell lysates were MAP kinase in these cells. EGF treatment at 100 ng/ml subjected to Western analysis with anti active MAP-kinase for the same time was a less potent activator of MAP antibodies kinase in these cells, while estrogen treated samples had undetectable levels of phosphorylated elk-1. Anti-active MAP kinase antibodies were also used to speci®cally detect tyrosine-phosphorylated erk1 and erk2 in MCF- truncated FGF receptor-2 (Bek 2.0 kb), which we have 7 cells treated with FGFs. A time course of MAP shown previously to have a dominant inhibitory kinase activation shows that the active MAP kinase activity on FGF signaling (Li et al., 1994). Expression signal is still seen after 5 h of treatment with 10 ng/ml of Bek 2.0 kb considerably decreased the FGF- FGF-4 and the presence of estrogen does not inhibition of CAT activity in the MCF-7 cells. appreciably appear to a€ect if (Figure 3b). MAP Expression of a construct encoding the wild type kinase is activated by FGF with similar kinetics in FGFR-2 reduced overall CAT activity, presumably due T47D cells (Figure 3c). The amount of protein in each to receptor activation resulting from high levels of lane is equal as can be seen in Figure 9 where the same expression (Li et al., 1997), but FGF-mediated blot is reused. inhibition of the estrogen response was still detect- able. Cotransfection with a plasmid encoding the activated FGFR-2/C382R strikingly reduced the E€ect of FGF on estrogen-induced gene expression activity even further although estrogen-inducibility Factors such as EGF which activate the MAP kinase was still maintained. In this case however, as pathway can activate the estrogen receptor in an expected, FGF treatment had no e€ect (Figure 4a). estrogen-independent fashion (Kato et al., 1995; To rule out the possibility that the e€ect was due to Bunone et al., 1996). We studied the e€ect of FGF a peculiarity of the estrogen receptor in MCF-7 cells, treatment on an ERE-CAT reporter (DETCO) con- we performed the same experiment in NIH3T3 cells struct transfected transiently in MCF-7 cells. The cotransfected with an ER expression plasmid (NIH3T3 DETCO has an ERE upstream of a minimal promoter cells do not express estrogen receptor endogenously). from the TK gene and has been previously used by Although FGF is a potent mitogen for ®broblasts, other researchers to study ER-mediated activation of FGF treatment caused a similar (®vefold) reduction of transcription (Bunone et al., 1996; Trowbridge et al., ERE-CAT activity in cells treated with estrogen 1997). As expected, the CAT activity was strongly (Figure 4b). The Bek 2.0 kb construct was unable to induced in cells treated with estrogen. Treatment with block the inhibitory e€ect of FGF in the NIH3T3 cells. 10 ng/ml FGF-4 had little e€ect on basal activity, but This is most likely due to the high level of endogenous reduced estrogen-induced CAT activity about 2.5-fold FGF receptors expressed on NIH3T3 cells that cannot from the level induced by estrogen alone. This be suciently blocked by the transfected truncated inhibitory e€ect of FGF was not observed in similar receptor. On the other hand, expression of the experiments where DETCO was replaced with a CAT activated FGFR-2 had an e€ect very similar to that reporter plasmid containing a TK promoter alone observed in MCF-7 cells. This repressive e€ect of FGF (data not shown). We also did a cotransfection with a on estrogen-stimulated ERE-CAT activity is not seen Inhibitory effects of FGF on MCF-7 cells MR Johnson et al 2651 in T47D cells (data not shown). In fact, in these cells essential for the growth factor-induced ligand-indepen- FGF was able to induce ER-mediated transcription in dent activity of ER (Kato et al., 1995; Bunone et al., an estrogen-independent manner (data not shown). 1996). We tested whether the inhibitory e€ect of FGF This e€ect is similar to that previously reported for on ER activity was mediated by phosphorylation at EGF and IGF in other cell types (Kato et al., 1995). this site. A mutant S118A ER was transfected into NIH3T3 cells along with the ERE-CAT reporter construct, and estrogen and FGF were applied as FGF can modulate the level of progesterone receptor before. Figure 6 shows that CAT activity induced by mRNA the S118A mutant ER is still inhibited by FGF Estrogen regulates the mRNA level of several genes that although the estrogen-inducibility is slightly reduced have an ERE in the promoter region. We looked at the compared to wild type ER. The level of ER transiently expression of progesterone receptor (PR) which is expressed in these cells was determined by Western blot upregulated by estrogen (Katzenellenbogen and Nor- to be approximately equal within each transfection man, 1990). Figure 5 shows that estrogen can induce the group (data not shown). expression of PR by 6 h in MCF-7 cells. FGF treatment clearly inhibits the estrogen e€ect and results in a lower p21/WAF-1 is induced by FGF in MCF-7 cells steady-state level of PR mRNA after 24 h. We looked for changes in the levels of known G1 cell cycle inhibitory proteins that may provide clues about The inhibitory e€ect of FGF is independent of MAP the nature of the FGF inhibitory e€ect. We found that kinase phosphorylation of ER Ser 118 the level of cyclin D1 in FGF-treated cells was The serine 118 residue in the N-terminal activation unchanged at 22 h but upregulated at about 5 h both domain of ER is a substrate for MAP kinase and is in the presence and absence of estrogen (Figure 7b), while the level of the cdk inhibitor p27/kip-1 was unchanged over 22 h. However, the cdk inhibitor p21/ a waf-1 was induced by the FGFs, particularly by FGF-2 (Figure 7a). Interestingly, the level of induction produced by di€erent FGFs correlated with the extent of the inhibitory e€ect, with FGF-2 being the strongest and FGF-7 being the weakest. p21 levels in clone 1021 #18, which expresses the activated FGF receptor, are much higher than in MCF-7 cells and are unchanged by FGF treatment (Figure 7a, right panel). T47D cells under the same conditions showed no detectable b induction of p21 by FGF (data not shown). Figure 7b shows that the level of p21 induced by FGF remains elevated for at least 22 h. The induction of p21 by FGF is also seen at the level of mRNA. In MCF-7 cells treated with FGF4 for 48 h, p21 mRNA is induced (Figure 8). The induction by FGF is seen at 1 h and is una€ected by the presence of estrogen (data not shown). There is no further induction of p21 by FGF treatment in 1021 #18 cells where the basal p21 level is much higher than in MCF- Figure 4 CAT assays. Cells were transiently transfected with a 7 parental cells. dominant negative truncated mutant (Bek 2.0 kb), murine FGF receptor-2 (Bek), an activated murine FGFR-2 mutant (1021), or vector pRK5, along with DETCO, a plasmid containing a CAT reporter gene under the control of an estrogen response element- containing promoter. Forty-eight hours later, each set of transfected cells was treated with the four conditions indicated (a) MCF-7 cells. (b) NIH3T3 cells co-transfected with wild type estrogen receptor (CMV-ER)

t=0 6 h 24 h

PR

GAPDH

50 nM E2: – ––+ +++ 50 ng/ml FGF4: –––++++ Figure 5 Northern blot analysis. MCF-7 cells were treated as indicated, and mRNA was prepared. Blotting was performed with Figure 6 CAT assay with wild type and S118A mutant ER. 3.1 kb BamHI fragment from the human progesterone receptor NIH3T3 cells were transfected with the ERE-CAT construct and (PR) and 0.6 kb PstI fragment from the rat glyceraldehyde either wild type estrogen receptor or the phosphorylation site phosphate dehydrogenase messages (S118A) mutant receptor, and treated as indicated Inhibitory effects of FGF on MCF-7 cells MR Johnson et al 2652

a – – E2+FGF4 E2+FGF4 E2 E2 FGF4 FGF4

p21 E2 No addition No addition FGF–1 FGF–2 FGF–4 FGF–7 E2+FGF–2 FGF–2 FGF–4 No addition (0.5%) FGF–2 (0.5%) FGF–2 (0.5%) waf–1 p21 GAPDH MCF7 1021#18 MCF–7 1021#18 Figure 8 p21 mRNA levels in MCF-7 and 1021 #18 cells. Cells b p21waf–1 were treated for 48 h with 10 nM estrogen, 10 ng/ml FGF4, or both. Total RNA was collected and subjected to Northern blotting using probes directed against the p21 and glyceraldehyde cyclin D1 phosphate dehydrogenase messages

p27 Time: 0 15'5 h22 h 5 h22 h 15' 5 h 22 h E2: – – – – + + ++ + FGF4: – +++–– +++ MCF–7 T47D

Figure 7 Western blots. (a) MCF-7 and 1021 #18 cells were Time 005'30'60'18 h30' 5' 30' 60' 18 h 30' treated overnight with 10 ng/ml FGFs or 50 ng/ml EGF or 10 nM Treatment ––FF F FII FF F F estradiol as indicated. Whole cell extracts were analysed for the p21/WAF1/CIP1 protein. (b) MCF-7 cells were treated with Phospho-Stat1 10 nM estradiol or 10 ng/ml FGF4 for the indicated times and Western blotting was performed using antibodies directed against Stat 1 p21, cyclin D1 and p27 Figure 9 MCF-7 and T47D cells were treated with 100 ng/ml FGF2 (F) or 150 ng/ml IFN-g (I) as a positive control and harvested at the indicated timepoints. Lysates were subjected to Western blotting using antibodies directed against Stat1 and STAT1 is phosphorylated upon FGF treatment of phosphorylated Stat1 MCF-7 cells The inhibitory e€ect of FGF on MCF-7 cells and the sustained induction of p21 that it causes, is unchar- Discussion acteristic in light of the mitogenic properties of FGF on many cell types. We reasoned that this e€ect might Breast tumors are a mix of epithelial, stromal and be via a signaling mechanism that is independent of the vascular components, but the interactions between ras-MAP kinase pathway. A signaling pathway these cell types are not well de®ned. Although the originally associated with growth arrest was the JAK- malignant cells are epithelial in origin, growth factors STAT pathway in which the JAK family of tyrosine provided by the stromal cells probably play an kinases phosphorylate and activate STAT proteins important role in regulating epithelial cell growth. (Darnell et al., 1994; Ihle and Kerr, 1995). STAT Expression of FGF-2 in human breast tissue was activation in response to EGF and interferon (IFN)-g consistent with this assumption since FGF-2 was found has been associated with growth inhibition and localized to the basement membrane and in the induction of p21 (Chin et al., 1996). We determined myoepithelial cells, but not in the luminal epithelial whether FGF treatment activated the STAT pathway cells (Gomm et al., 1991). MMTV-induced mouse in MCF-7 cells. Figure 9 shows a Western blot mammary tumors are thought to arise by autocrine performed with antibodies directed speci®cally against stimulation of mammary epithelial cells due to phosphorylated STAT1. FGF treatment induces a very activation of growth factor gene expression by viral signi®cant level of phosphorylation of STAT1 within integration (Peters, 1991). In transgenic animals 5 min, which decreases progressively with time but is expressing int-1, infection with MMTV leads to a still detectable at 18 h after FGF addition. As expected high incidence of mammary carcinomas. These tumors STAT1 phosphorylation is strongly induced by IFN-g often express int-1 together with FGF-3 or FGF4, treatment. On the other hand, induction of STAT1 suggesting that these factors act in cooperative ways to phosphorylation by FGF in T47D cells is barely promote mammary carcinogenesis in mice (Shackleford detectable, while IFN-g can still strongly induce its et al., 1993; MacArthur et al., 1995). In humans, the phosphorylation. The lower panel is a Western blot response to these growth factors may normally be using antibodies directed against STAT1. This result directly or indirectly under hormonal control. As was con®rmed by immunoprecipitation with anti- tumors progress, they often become hormone-indepen- STAT1 antibodies and immunoblotting with anti- dent and therefore resistant to anti-estrogenic thera- phosphotyrosine antibodies (data not shown). We did pies. Genetic or epigenetic events leading to not detect phosphorylation of other STAT molecules. unregulated activation of growth factor signaling Immunoprecipitation with anti-STAT2, STAT3, or pathways probably enables the tumor cells to bypass STAT5 antibodies and anti-phosphotyrosine immuno- or be less responsive to hormonal control and thus blotting revealed no activation of these STATs by FGF become more malignant. This appears to be the case in in MCF-7 cells (data not shown). the Dunning rat prostate tumor model where FGFs Inhibitory effects of FGF on MCF-7 cells MR Johnson et al 2653 and FGFR-2 genes are activated in malignant parental MCF-7 cells is the ability of the FGF- epithelial cells, an event which presumably frees them expressing cells to form tumors in vivo, where the from dependence on stromally-derived FGF-7 (Yan et angiogenic ability of FGFs could play a role. It is also al., 1993). interesting to note that MDA-MB-134 cells which We were originally interested in determining what overexpress FGFR-1 and maintain an estrogen factors were involved in converting cells from an response, are growth inhibited by FGF (McLeskey et estrogen-dependent state to an estrogen-independent al., 1994). one. Based on the prostate model, we originally set out It is now also becoming increasing clear that the to test the hypothesis that autocrine FGF production FGF e€ect on breast epithelial cells is not like that of by mammary epithelial cells enabled them to bypass EGF i.e., FGF does not activate ER in a ligand- the requirement for estrogen. However, inhibition of independent manner. Studies of FGF-1 and FGF-2 FGF signaling in the estrogen-unresponsive HBL100 expression in breast tumor biopsies showed that they breast epithelial cell line, which produces FGF-2, did are expressed at lower levels in tumors (Bansal et al., not restore estrogen responsiveness (our unpublished 1995; Yiangou et al., 1997). The loss of FGF-2 results). Furthermore, when we expressed constitutively expression correlated with poor prognosis as well, activated FGF receptors in the hormone dependent pointing to a negative regulatory role for FGF in MCF-7 cells, expecting that we could bypass the breast tumor development. Also, expression of FGFs hormone requirement by activating a constitutive has not been found to be elevated in mammary tumors growth factor signaling pathway, we found that the that are not MMTV-induced (Coleman-Krnacik and clones expressing the activated FGF receptors grew Rosen, 1994). Fenig et al. (1997) have recently reported poorly and still required estrogen for growth. We a ®nding similar to ours; that FGF-2 has a growth therefore set out to investigate this ®nding by inhibitory e€ect on MCF-7 cells in spite of activating comparing the e€ects of FGF signaling in the ER- MAP kinase but the e€ect on the estrogen response has positive breast cell line MCF-7, and in the T47D cell not previously been described. line, which is estrogen-responsive but expresses ER at a It is unlikely that MAP kinase is mediating the much lower level. inhibitory e€ect of FGF since all the FGFs tested are As described in this report, we were able to show able to strongly activate MAP kinase while having that FGF signaling has an inhibitory e€ect on the di€erent inhibitory e€ects, and T47D cells respond to growth of MCF-7 cells with an increased frequency of FGF treatment with MAP kinase activation, but are cells in G1 phase of the cell cycle. FGF also has a not growth inhibited and do not upregulate p21 repressive e€ect on estrogen-induced transcription as expression. Also, the major MAP kinase target, serine seen by the inhibition of ERE-CAT activity and by 118 of ER, does not appear to be involved in the Northern analysis of PR, an estrogen-regulated gene. inhibitory e€ect on the estrogen-inducible CAT The inhibitory e€ect appears to be mediated by the cdk activity in NIH3T3 cells. In PC12 cells, the decision inhibitor p21 which is strongly induced by FGF. T47D to proliferate or di€erentiate is thought to be made cells are not growth inhibited by FGF, show no based on the amplitude and duration of the MAP inhibition of an estrogen-inducible CAT reporter, and kinase signal (Marshall, 1995), and in NIH3T3 cells no induction of p21. Increasing the level of FGF expressing trkA, NGF inhibits growth via activation signaling by overexpression of an activated FGFR-2 of the MAP kinase pathway (Pumiglia and Decker, (1021) strongly inhibited cell growth and the expression 1997). Use of the mek-1 inhibitor PD98059 did not of the ERE-CAT plasmid. Cells expressing FGFR-2/ abolish the growth-inhibitory e€ect of FGF, but it is 1021 also had elevated levels of p21 and activated possible that the residual active MAP kinase is MAP kinase in the absence of exogenous FGF sucient for inhibition (data not shown). Thus the treatment, indicating that the activated receptor was possibility that the duration of the MAP kinase signal functional in these cells. Thus FGF and estrogen have is regulating the inhibitory e€ect cannot be entirely opposing functions in MCF-7 cells and FGFs may ruled out. have a role in modulating the estrogen response and While our results clearly show a dual e€ect of FGF epithelial cell growth and in breast tissues. signaling in MCF-7 cells, inhibition of growth and These results were somewhat surprising in view of down-regulation of estrogen receptor-mediated gene the hypothesis that autocrine growth factor loops can expression, the question of whether these e€ects are relieve the need for estrogen stimulation in breast independent of each other or mediated by the same cancer cells. The data in the literature about the e€ects primary e€ector remains to be answered. of FGFs on breast epithelial cells is, however, We found that FGF induces phosphorylation of con¯icting. As in the prostate model, it was assumed STAT1 and elevates levels of p21, and it is likely that that stromally derived FGFs stimulate the growth of the negative e€ect on the growth rate is mediated by epithelial cells. Acquisition of an autocrine FGF p21 activity. p21 can negatively regulate the activity of growth loop would then free the cells from stromal several cyclin-cdk complexes (Sherr and Roberts, control during tumor progression. Supporting this 1995). One such G1 cyclin, cyclin D1, has been shown hypothesis is the ®nding that certain ER-negative to be essential for proliferation of breast epithelium; breast tumor-derived cell lines express some of the mice lacking cyclin D1 show a defect in proliferation of FGFs (Dickson et al., 1986; Ethier, 1995); and MCF-7 the estrogen-responsive breast epithelium during cells transfected with a cDNA encoding FGF-4 formed pregnancy (Sicinski et al., 1995). Wang et al. (1997) tumors in ovariectomized mice (McLesky et al., 1993). have recently also reported an FGF-induced elevation These latter data contrast with our ®ndings, but it of p21 in MCF-7 cells which correlated with a should be noted that the major di€erence found by disappearance of active cdk2, (a cyclin D1 partner) these authors between FGF-4-expressing MCF-7 and and dephosphorylation of Rb. These authors also Inhibitory effects of FGF on MCF-7 cells MR Johnson et al 2654 found an increase in p21-associated PCNA, cyclin D1, Materials and methods and cyclin E, strongly indicating that the p21 increase is at least partly responsible for the growth inhibition. Cell lines p21 expression is transiently induced by growth factors MCF7 and T47D are estrogen-responsive breast cancer cell in cycling cells but does not lead to cell cycle arrest lines. BT549, MDA-MB-231 and H578T are estrogen (Michieli et al., 1994). On the other hand, sustained receptor-negative tumor lines. HBL100 and H578Bst are expression of p21 (as is seen in MCF-7 cells with FGF derived from normal mammary epithelial cells. These lines treatment) is associated with growth inhibition (Halevy were obtained from ATCC and were maintained in et al., 1995). A recent report by Su et al. (1997) shows Dulbecco's Modi®ed Eagle's Medium (DMEM) supple- that a constitutively activated FGFR-3 can activate mented with 10% fetal calf serum (FCS), and 100 units/ml STAT1, leading to an upregulation of p21 and cell penicillin and 100 mg/ml streptomycin (P/S). NIH3T3 cells are a mouse ®broblast cell line grown in DMEM cycle arrest. EGF induces STAT1 only in A431 cells supplemented with 10% calf serum. 1021 #18 cells were where it is growth inhibitory, and a link between STAT obtained by calcium phosphate transfection of MCF7 cells activation and p21 induction has been established with a plasmid containing the activated murine FGFR-2 (Chin et al., 1996). Thus the inhibitory e€ect of FGF mutant C382R in the pRK5 vector (Li et al., 1997), and on growth in MCF-7 cells is likely to be via a similar selection by neomycin resistance. 1021 #18 cells were mechanism. maintained in DMEM+10% FCS+400 mg/ml Geneticin A possible mechanism by which elevated p21 might (Life Technologies, Grand Island, NY). also lead to the down-regulation of the estrogen response is suggested by some recent ®ndings. Zwijsen Reverse-transcriptase PCR et al. (1997) showed that cyclin D1 was able to bind cDNA was synthesized from 2 mg heat-denatured total ER in the absence of a cdk partner to activate ER- RNA in 20 ml total volume. PCR was performed using 1 ml mediated transcription and that the binding of cyclin of the reverse-transcription reaction product in a 40 ml D1 to ER is via the hormone binding domain of ER. reaction containing 1 U TAQ polymerase (Perkin Elmer, Furthermore, Trowbridge et al. (1997) have recently Foster City, CA) in reaction bu€er plus 200 mM dNTPs,

shown that the cycA/cdk2 complex can phosphorylate 200 ng of each primer, and 1.5 mM MgCl2.Thirty-®ve ER and enhance its transcriptional activity. Therefore, cycles of 45 s at 948C, 1 min at 558C, and 2 min at 728C it may be possible that p21 inhibits ER-mediated were followed by a ®nal extension of 9 min at 728C. transcription either by blocking cyclin/cdk phosphor- Products were electrophoresed on 1% agarose gels, ylation of ER, or by sequestering cyclin D1 in a transferred to nylon ®lters and subjected to Southern blot 32 complex that prevents it from binding ER. We have analysis using the appropriate P-labeled random-primed probes as described previously (Ittman and Mansukhani, looked at the e€ect of increasing p21 levels on the 1997), and exposed for autoradiography. estrogen-induced CAT activity in MCF-7 and NIH3T3 cells by cotransfecting varying amounts of a p21- encoding expression plasmid. Preliminary results Flow cytometry indicate that higher p21 levels cause a modest Cells in 6 cm dishes were kept in phenol red-free DMEM inhibition of the estrogen response (data not shown). (Mediatech) supplemented with 5% charcoal-stripped fetal Although it may be possible that the mechanisms for calf serum. Treatments were applied for 24 h after which growth inhibition and down-regulation of the estrogen the cells were trypsinized and spun in a microfuge. The cell response are di€erent, they may both be mediated pellets were washed with phosphate-bu€ered saline (PBS), through di€erent activities of p21. The activation of resuspended in 1 ml and transferred to 15 ml Falcon tubes. Fixing was done with the addition of 4 ml of ice-cold 80% the STAT-1 pathway and sustained p21 induction has ethanol with simultaneous vortexing and overnight not been generally observed following stimulation of incubation at 48C. Cells were recovered by centrifugation, FGF receptor signaling (Silvennoinen et al., 1993). pellets were air-dried and resuspended in 800 mlPBS This is perhaps not surprising, as FGF signaling has containing 200 mg/ml RNase A for 2 h at 378C. Cells been studied mainly in the context of mitogenic were stained with 0.2 ml 56propidium iodide (200 mg/ml induction, but it does raise the question of why FGF propidium iodide, 40 mM sodium citrate, pH 7) for 1 h and would activate a signal transduction pathway in MCF- DNA content was analysed on a FACScan ¯ow cytometer. 7 cells that is not seen in other cells. While super®cially The percentage of cells in G0/G1, S and G2/M was our data could be consistent with the hypothesis that determined by ModFit software. the presence of ER in¯uences the mode of FGF signaling, this hypothesis is at present totally spec- CAT assays ulative. Another possibility is suggested from the recent MCF-7 and NIH3T3 cells were plated at 1.56106 cells/ ®ndings that signaling by a mutant FGFR-3 can plate on 10 cm plates and allowed to attach overnight. The induce STAT-1 activation leading to p21 induction following day, plates received fresh DMEM+10% FCS, (Su et al., 1997). While our data suggest that this may and were transfected several hours later by calcium not be exclusively related to FGFR-3 (our 1021 phosphate. All transfections were performed with 10 mg construct is a mutant of FGFR-2) it could be thought DETCO (a plasmid containing the chloramphenicol that this pathway is activated by speci®c FGF receptor acetyltransferase gene under the control of an ERE heterodimers, perhaps including FGFR-3. It is inter- inserted upstream of a minimal thymidine kinase promoter esting to note that MCF-7 cells express ®ve types of (7109) and 3 mgCMV-bgal (a plasmid constitutively expressing the b-galactosidase gene under the control of FGF receptor isoforms, thus permitting a variety of the CMV promoter) as an internal control. Where combinations of receptor heterodimers, while fibro- indicated, the plates also received either 8 mg of plasmid blasts generally express only FGFR-1 and FGFR-2 (Li encoding murine wild type FGFR-2, or equimolar amounts et al., 1994). We are currently investigating this of plasmid 1021 encoding activated murine FGFR-2, or hypothesis. dominant negative FGFR-2, or pRK5 vector. NIH3T3 Inhibitory effects of FGF on MCF-7 cells MR Johnson et al 2655 cells were also cotransfected with 10 mg of plasmid reactions were subjected to SDS ± PAGE, and immuno- encoding either the wild type human estrogen receptor or blotted using phospho-speci®c Elk-1 antibodies. the S118A mutant under the control of the CMV promoter. Sixteen hours after application of calcium Western blots and antibodies phosphate precipitate, plates were washed with PBS and fed fresh phenol red-free DMEM+5% charcoal-stripped Protein samples were subjected to SDS ± PAGE and fetal calf serum (CS-FCS). Cells were washed and refed transferred to nitrocellulose ®lters. Western blot analysis three times over the next 8 h, and treated overnight with was performed according to standard protocols and 10 nM 17-b-estradiol, 10 ng/ml FGF4, or both. Cells were developed by enhanced chemiluminescence (Amersham). harvested, resuspended in 250 mM Tris, pH 7.4, and lysed Anti-p21, anti-p27, and anti-cycD1 (mouse monoclonal by ®ve cycles of freezing and thawing. Clari®ed lysates antibodies) were from Transduction Laboratories (Lex- were used in CAT assays using 14C-chloramphenicol and ington, KY). Anti-ER (HC-20, rabbit polyclonal) and anti- analysed by thin layer chromatography. Fractional ERK2 (C-14, rabbit polyclonal) were from Santa Cruz conversion was calculated by phosphorimager analysis, Biotechnology (Santa Cruz, CA). Anti-active MAP kinase adjusted by b-galactosidase assay readings, and normalized polyclonal antibody was from Promega (Madison, WI). to the fractional conversion of untreated cells to yield Anti phospho-Stat-1 was from New England Biolabs relative CAT activity. (Beverley, MA). Anti-Stat-1 rabbit polyclonal antibodies were obtained from David Levy, NYU Medical Center. Sheep anti-mouse and donkey anti-rabbit secondary Tritiated thymidine incorporation antibodies, and enhanced chemiluminescence reagent were Cells that had been growing for 1 week in phenol red-free from Amersham (Arlington Heights, Illinois). DMEM supplemented with 5% CS-FCS were plated at 36104 cells per well in 24-well plates. The following day, Northern blots cells were starved in phenol red-free DMEM+0.5% CS- FCS. Two days later the cells were supplemented with 0, 1, Estrogen-deprived MCF-7 or 1021 #18 cells were treated as or 10 nM estradiol, and 0, 1, or 10 ng/ml FGF4 and indicated mRNA was puri®ed from 250 mgoftotalRNA incubated overnight. Cells were labeled for 6 h with 1 mCi/ using the Oligotex mRNA kit (Qiagen, Chatsworth, CA). ml 3H-labeled thymidine, and treated with 10% trichlor- Samples were run out on formaldehyde-containing 1% oacetic acid at 48C overnight, washed, and lysed in 0.5 M agarose gels and transferred to Genescreen ®lters, which NaOH. Samples were neutralized with 0.5 M HCl, added to were hybridized using the appropriate probes and imaged scintillation cocktail, and counted. by autoradiography and/or phosphorimager (Molecular Dynamics). In vitro kinase assays In vitro MAP kinase assays were performed using the MAP Kinase Assay Kit according to the manufacturer's instructions (New England Biolabs, Beverly, MA). Acknowledgements Brie¯y, 26106 MCF7 cells were plated on 10 cm plates, We would like to thank Michael J Garabedian for serum-starved and estrogen-deprived overnight, treated for supplying CMV-ER, CMV-ER(S118A) and DETCO plas- 30 min with 100 ng/ml FGF2, FGF4, or EGF, or 100 nM mids, as well as expert advice at all stages of this work, and estradiol, harvested and lysed. Clari®ed lysates were David E Levy for anti-STAT-1 antibodies. This research immunoprecipitated using phospho-speci®c MAP kinase was supported by PHS grant CA42568 from the National antibodies. Kinase reactions included the immunoprecipi- Cancer Institute (CB) and IRG 14-37 from the Kaplan tates and elk-1 as a substrate for MAP kinase. The Cancer Center (AM).

References

Adnane J, Gaudray P, Dionne CA, Crumley G, Jaye M, Dickson RB and Lippman ME. (1995). Endocr. Rev., 16, Schlessinger J, Jeanteur P, Birnbaum D and Thiellet C. 559 ± 581. (1991). Oncogene, 6, 659 ± 663. Dickson RB, McManaway ME and Lippmann ME. (1986). Altucci L, Addeo R, Cicatiello L, Dauvois S, Parker MG, Science, 232, 1540 ± 1543. Truss M, Beato M, Sica V, Bresciani F and Weisz A. Ethier SP. (1995). J. Natl. Cancer Inst., 87, 964 ± 972. (1996). Oncogene, 12, 2315 ± 2324. Fenig E, Wieder R, Paglin S, Wang H, Persaud R, Bansal GS, Yianou C, Coope RC, Gomm JJ, Luqmani YA, Haimovitz-Friedman A, Fuks Z and Yahalom J. (1997). Coombes RC and Johnston CL. (1995). Br. J. Cancer, 72, Clin. Cancer Res., 3, 135 ± 142. 1420 ± 1426. Gomm JJ, Smith J, Ryall GK, Baillie R, Turnbull L and Basilico C and Moscatelli D. (1992). Adv. in Cancer Res., 59, Coombes RC. (1991). Cancer Res., 51, 4685 ± 4692. 115 ± 165. Halevy D, Novitch BG, Spicer DB, Skapek SX, Rhee J, Bunone G, Briand PA, Miksicek RJ and Picard D. (1996). Hannon GJ, Beach D and Lassar AB. (1995). Science, 267, EMBO J., 15, 2174 ± 2183. 1018 ± 1021. Chalbos D, Philips A and Rochefert H. (1994). Sem. Cancer Horwitz KB, Zava DT, Thilagar AK, Jensen EM and Biol., 5, 361 ± 368. McGuire WL. (1978). Cancer Res., 38, 2434 ± 2437. Chin YE, Kitagawa M, Su W-CS, You ZH, Iwamoto Y and Ignar-Trowbridge DM, Nelson K, Bidwell MC, Curtis SW, Fu X-Y. (1996). Science, 272, 719 ± 722. Washburn TF, McLachlan JA and Korach KS. (1992). Coleman-Krnacik S and Rosen JM. (1994). Mol. Endocri- Proc. Natl. Acad. Sci. USA, 89, 4658 ± 4662. nol., 8, 218 ± 229. Ihle JN and Kerr IM. (1995). Trends Genet., 11, 69 ± 74. Curtis SW, Washburn T, Sewall C, DiAugustine R, Lindzey Ittman M and Mansukhani A. (1997). J. of Urol,, 157, 351 ± J, Couse JF and Korach KS. (1996). Proc. Natl. Acad. Sci. 356. USA, 93, 12626 ± 12630. Johnson DE and Williams LT. (1993). Adv. Cancer Res., 60, Darnell Jr JE, Kerr IM and Stark GR. (1994). Science, 264, 1±41. 1415 ± 1421. Inhibitory effects of FGF on MCF-7 cells MR Johnson et al 2656 Kato S, Endoh H, Masuhiro Y, Kitamoto T, Uchiyama S, Peters G. (1991). Sem. Virol., 2, 319 ± 328. Sasaki H, Masushige S, Gotoh Y, Nishida E, Kawashima Pumiglia KM and Decker S. (1997). Proc. Natl. Acad. Sci., H, Metzger D and Chambon P. (1995). Science, 270, 94, 448 ± 452. 1491 ± 1494. Shackleford GM, MacArthur CA, Kwan HC and Varmus Katzenellenbogen BS and Norman MJ. (1990). Endocrinol- HE. (1993). Proc. Natl. Acad. Sci., 90, 740 ± 744. ogy, 126, 891 ± 898. Sherr CJ and Roberts JM. (1995). Genes Dev., 9, 1149 ± 1163. Li Y, Basilico C and Mansukhani A. (1994). Mol. Cell. Biol., Sicinski P, Donaher J, Parker S, Li T, Fazeli A, Gardner H, 14, 7660 ± 7669. Haslam S, Bronson R, Elledge S and Weinberg R. (1995). Li Y, Mangasarian K, Mansukhani A and Basilico C. (1997). Cell, 82, 621 ± 630. Oncogene, 14, 1397 ± 1406. Silvennoinen O, Schindler C, Schlessinger J and Levy DE. Luqmani YA, Bansal GS, Mortimer C, Buluwela L and (1993). Science, 261, 1763 ± 1739. Coombes RC. (1996). Eur. J. Cancer, 32, 518 ± 524. Su W-C, Kitagawa M, Xue N, Xie B, Garofolo S, Cho J, MacArthur CA, Shankar DB and Shackleford GM. (1995). Deng C, Horton W and Fu X-Y. (1997). Nature, 386, J. Virol., 69, 2501 ± 2507. 288 ± 292. Marshall CJ. (1995). Cell, 80, 179 ± 185. Trowbridge JM, Rogatsky I and Garabedian MJ. (1997). McLeskey SW, Ding IY, Lippman ME and Kern FG. (1994). Proc. Natl. Acad. Sci., 94, 10132 ± 10137. Cancer Res., 54, 523 ± 530. Tsai MJ and O'Malley BW. (1994). Ann. Rev. Biochem., 63, McLeskey SW, Kureybayashi J, Honig SF, Zwiebel JZ, 451 ± 486. Lippman ME, Dickson RB and Kern FG. (1993). Cancer Wang H, Rubin M, Fenig E, DeBlasio A, Mendelsohn J, Res., 53, 2168 ± 2177. Yahalom J and Wieder R. (1997). Cancer Res., 57, 1750 ± Michieli P, Chedid M, Lin D, Pierce J, Mercer WE and Givol 1757. D. (1994). Cancer Res., 54, 3391 ± 3395. Yan G, Fukabori Y, McBride G, Nikolaropolous S and Muller WJ, Frederick SL, Dickson C, Peters G, Pattengale P McKeehan WL. (1993). Mol. Cell. Biol., 13, 4513 ± 4522. and Leder P. (1990). EMBO J., 9, 907 ± 913. Yiangou C, Gomm JJ, Coope RC, Law M, Luqmani YA, Murakami A, Tanaka H and Matsuzuwa A. (1990). Cell Shousha S, Coombes RC and Johnston CL. (1997). Br. J. Growth Di€., 1, 225 ± 231. Cancer, 75, 28 ± 33. Ornitz DM, Xu J, Colvin JS, McEwen DG, MacArthur CA, Zwijsen RML, Wientjens E, Klompmaker R, van der Sman Coulier F, Gao G and Goldfarb M. (1996). J. Biol. Chem., J, Bernards R and Michalides RJAM. (1997). Cell, 88, 271, 15292 ± 15297. 405 ± 415. Payson RA, Wu J, Liu Y and Chiu I-M. (1996). Oncogene, 13, 47 ± 53.