Members of the fibroblast growth factor (FGF) family, which normally control cerebellar neuronal maturation, may represent more natural and less toxic tools with which to target medulloblastoma (MB), an embryonal brain tumor thought to arise from cerebellar neuronal precursors. In support of this, we found previously basic FGF/FGF-2 can inhibit MB progression by inducing neuronal-like differentiation, slowing the growth, and triggering apoptosis of a MB cell line we established from a histopathologically classic tumor (R. L. Kenigsberg et al., Am. J. Pathol., 151: 867–881, 1997). In the present study, the usefulness of this approach is additionally investigated. We report that of the five FGFs found in the developing cerebellum, only two, FGF-2 and FGF-9, possess antitumoral activity for MB. This activity is only noted for cell lines that originate from classic (UM-MB1 and SYR) rather than desmoplastic (HSJ) tumors. Whereas these FGFs inhibit proliferation of both classic cell lines, they only advance neuronal differentiation and induce apoptosis of one, UM-MB1. Consistent with these responses, after FGF treatment, levels of neurofilaments and the proapoptotic modulator Bax only increase in UM-MB1, whereas the cyclin-dependent kinase inhibitor p27/Kip1 (p27), which accumulates in cerebellar neuronal precursors before they exit the cell cycle, goes up in both UM-MB1 and SYR. Finally, although the histological variant of MB may help predict the sensitivity of MB to the FGFs, the selectivity, specificity, and type of response elicited may be dictated by, as evident by immunoprecipitation and Western blot analyses, the expression of certain FGF receptor types.
MB,3 a highly aggressive embryonal tumor of the cerebellum that occurs most frequently in childhood, accounts for close to 20% of all of the intracranial tumors in the pediatric population (1). Despite marked improvements in surgical, imaging, and radiation techniques, as well as the introduction of adjuvant chemotherapy in its management, the high incidence of MB for recurrence and proclivity for dissemination render its long-term clinical picture consistently poor (2).
Although little is known about the transforming events that underlie the development of this neoplasm, MB is thought to originate from two distinct germinal zones of the cerebellar cortex namely: b.f.(1) the earlier forming VM (3), which gives rise to the deep cerebellar neurons, Purkinje cells, neurons of the molecular layer as well as glia and ependyma (4), or b.f.(2) a later-forming matrix located to the outer surface of the cerebellum, the EGL of Obersteiner (5, 6), that gives rise exclusively to internal granule neurons (7). The differential localization of some antigens (8, 9) and receptors, like the neurotrophin receptor p75NTR (10), which distinguishes neuronal progeny of these two germinal zones to the two most common variants of MB, suggest that the classic variant originates from the VM, whereas the desmoplastic arises from the EGL.
Regardless of the precise histogenesis of MB, its derivation from the developing cerebellum is amply supported in studies that show this tumor to express distinct differentiation and lineage-related antigens found in cerebellar neuronal progenitors (11, 12, 13, 14, 15, 16, 17, 18). Expression of these markers in MB suggests that this tumor can, if appropriately stimulated, reinstate a more normal type of differentiation program to generate more mature and less aggressive progeny. To initiate this event, growth factors normally found in the developing cerebellum may be the most appropriate stimuli.
Unlike select neurotrophins that specifically impact the survival and differentiation of internal granule neurons, progeny of the EGL (19, 20), the FGFs, could, in view of their more widespread distribution in the developing cerebellum (21, 22, 23), influence cells of both germinal zones. In this regard, at least five members of the FGF family namely, acidic FGF/FGF-1 (24), bFGF/FGF-2 (25, 26), FGF-5 (27), FGF-6 (23), and FGF-9 (28, 29, 30) have been found in the developing cerebellum. Expression of these growth factors and their signaling tyrosine kinase receptors (FGFR1-FGFR4) change in a temporally and regionally specific fashion during cerebellum maturation, suggesting that they play very distinct roles within (22, 23, 31). To date, physiological roles for only one, FGF-2, in promoting cerebellar neurogenesis (32, 33) and aspects of neuronal differentiation (34, 35) have been evidenced. However, because levels of FGF-2 peak in the developing cerebellum before the generation of the internal granule neurons (25), it is likely that FGF-2 is more important to the maturation of the neurons that originate from the VM, the matrix from which the classic and most common variant of MB is thought to arise (9, 36).
In support of the above, our previous study shows FGF-2 can promote the maturation both in vitro (37) and in vivo (38), of a new MB cell line that we developed from a histopathologically classic tumor. However, this is only one cell line from one variant of MB, and FGF-2 is only one member of this growth factor family that localizes to the developing cerebellum. Therefore, in the present study, we extended these initial promising observations by developing and characterizing new cell lines from classic and desmoplastic tumors and by testing all five of the aforementioned FGFs for differentiating, cytostatic, and apoptotic activities. By following changes in the expression of important intracellular modulators of cell cycle, survival, and differentiation, we hoped to elucidate the underlying mechanism(s) by which the FGFs suppress the growth of this tumor. In addition, this comparative study was designed to identify markers, such as specific receptors, that are expressed on tumor cells of which the progression can be stopped by the FGFs.
MATERIALS AND METHODS
Characterization of Original Cerebellar Tumors.
Surgical specimens obtained at craniectomy from two intrinsic MB tumors located to the posterior fossi of two- and three-year-old males were either snap-frozen or formalin-fixed (3.7%) and paraffin-embedded for the immunohistochemical and histochemical analyses, or placed immediately in saline for preparing the two new continuous MB cell lines, HSJ and SYR. All of the tumors were classified according to the guidelines of the WHO for brain tumors (39). Routine histological stains used include H&E and Gomori for the detection of reticulin fibers. Immunohistochemical analyses of the tumors for the presence of lineage-related antigens like synaptophysin, NSE, S-100 protein, GFAP, vimentin, smooth and striated muscle actin, and cytokeratin were done on 2–5 μm thick sections using the peroxidase-immunoperoxidase detection technique. Endogenous peroxidase activity was routinely quenched with 3% H2O2 pretreatment of the tissues.
Establishment and Maintenance of MB Cell Lines.
For the establishment of the two new cell lines, SYR and HSJ, tumor specimens were minced and mechanically dissociated by mild trituration under aseptic conditions and the continuous lines then selected by subsequently passing proliferating cells. The two new adherent cell lines generated from them, HSJ and SYR, as well as our MB cell line characterized previously, UM-MB1 (37), were all maintained in logarithmic phase growth at 37°C, 5% CO2 in a humidified incubator in DMEM (Life Technologies, Inc., Burlington, Ontario, Canada) supplemented with 10 mm HEPES (pH 7.4; Life Technologies, Inc.), 10% FBS (Life Technologies, Inc.), and penicillin-streptomycin (Sigma Chemical Co., Oakville, Ontario, Canada). These predominantly adherent cultures were subcultured weekly by trypsinization with 0.125% in Ca2+, Mg2+-free HBSS (Life Technologies, Inc.) for 5 min (37). Doubling times for the new cells lines, HSJ and SYR, were determined during their exponential growth phases, which are maintained for several days when they are seeded at an initial density of 2 × 104 or 5 × 104 cells/cm2, respectively. Stock cells from passages 8–20 of the two new lines were expanded for freezing, and all of the experiments were performed within 10 passages from these stocks. The cell lines used in this study were free from Mycoplasma contamination as determined by routine testing with the enzyme immunoassay kit for cell culture (Boehringer Mannheim, Laval, Quebec, Canada).
Growth Factor Treatment, Viability, and Proliferation Assays.
To study the effects of the FGFs, equimolar (6.25 pm–6.25 nm) concentrations of human recombinant acidic FGF/FGF-1, bFGF/FGF-2, FGF-5, FGF-6, and FGF-9, (R & D Systems Inc., Minneapolis, MN) were added to all three of the MB cell lines several hours after seeding. For the viability and proliferation assays, cells were plated directly in 24 multiwell dishes (Falcon; Becton Dickinson Labware, Bedford, MA). Viable and nonviable cell counts were determined manually using a Neubauer hemocytometer and trypan blue exclusion. Changes in rates of proliferation after FGF treatment were assessed by measuring [3H]thymidine incorporation. For these measurements, cells were incubated with 0.5 μCi methyl-[3H]thymidine/well (25 Ci/mmol; Amersham Pharmacia Biotech., Baie d’Urfé, Quebec, Canada) for 12 h at 37°C (40) in replicate on various culture days. Wells were subsequently washed three times with ice-cold PBS, and trichloroacetic acid (10%) precipitable radioactivity solubilized in 0.1 n NaOH/1% SDS was then measured by liquid (Aquasol-2, Mississauga, Ontario, Canada) scintillation spectrometry.
Immunocytochemical Detection of Lineage-related Antigens.
For immunocytochemistry, cells were seeded in 24 multiwells containing poly-l-lysine-coated 12-mm diameter round glass coverslips. The presence of specific antigens was assessed on formalin-fixed (3.7%) cultures as detailed previously (37). Primary antibodies used at the indicated working dilutions include rabbit primaries such as: antibovine NSE (1:300; Dako, Dimension Labs, Mississauga, Ontario, Canada), anti-Mr 68,000 NF (NF-L; 1:200), anti-Mr 150,000 NF (NF-M; 1:400), anti-Mr 200,000 NF (NF-H; 1:400; Chemicon, Temecula, CA), anti-bovine GFAP (1:350; Dako), and the following mouse monoclonals: antibovine synaptophysin clone Sy35 (1:20; Dako), antivimentin clone V9 (1:50; Dako), anti-MAP1 clone HN-1, anti-MAP2 clone HN-2, and anti-MAP1b (MAP5) clone AA6 (all at 1:400; Sigma Chemical Co.). Immunoreactive sites were revealed using the Cy3-conjugated goat antibodies raised against rabbit or mouse IgG (1:1000; Jackson Research Labs, West Grove, PA). Fluorescent observations were recorded on T-MAX 400 ASA film (Kodak, Rochester, NY). Specificity of the immunoreactions was determined routinely with omission of the primaries and inclusion of appropriate preimmune sera.
In Situ Detection and Quantitation of Apoptosis.
UM-MB1 were seeded in T-25 flasks and maintained in log phase growth in the absence (control) or presence of equimolar (0.625 μm) of FGF-2 or FGF-9 for 7 culture days. Cells were harvested by trypsinization and pellets washed twice by centrifugation in PBS before formalin fixation. Fixed cells were washed twice in PBS and washed cell pellets resuspended to a final density of 2 × 106 cells/100 μl PBS. Cell suspensions (10 μl) were air dried onto glass slides and apoptotic cells detected by the Apoptag in situ indirect fluorescent detection kit according the manufacturer’s instructions (Intergen, Purchase, NY). Rhodamine-conjugated antidigoxigenin was used as the fluorochrome for detection of apoptotic nuclei and Ho 33342 (10 μg/ml; Sigma Chemical Co.) used as counterstain. Samples were coverslipped using Permount (Fisher Scientific, Nepean, Ontario, Canada) and stored in the dark at −20°C until microscopic examination. Quantitation of immunoreactive (apoptotic) and total (Ho) nuclei were done by counting stained cells in randomly distributed visual fields corresponding to ∼50% of the total coverslip area. Replicates (n = 9) of each condition were processed and counted for statistical analyses.
Western Blot Analyses.
Exponentially growing MB cells maintained in the absence (control) or presence of 0.625 nm FGF-2 or FGF-9 were harvested at various time intervals by trypsinization and washed twice with PBS. Cell extracts were prepared in lysis buffer [137 mm NaCl and 20 mm Tris (pH 8.0); Ref. 41] containing 1% NP-40, 10% (V/V) glycerol, and 0.1% SDS and the following protease inhibitors: leupeptin (0.2 μg/ml), aprotinin (5 μg/ml), and phenylmethylsulfonyl fluoride (1 mm). Protein contents of the cell extracts were determined using the Bio-Rad DC protein assay (Bio-Rad Laboratories, Mississauga, Ontario, Canada). For comparative purposes, equal amounts of cell homogenate protein (50 μg/well) were separated by SDS polyacrylamide (7.5% for NF; 12% for Bax, Bcl-2, Bcl-xL, or p27) gel electrophoresis then transferred electrophoretically onto Immobilon P membranes (Millipore, Nepean, Ontario, Canada). Equal protein loading was confirmed by Ponceau red staining. After washing out the stain, membranes were blocked for 1 h at room temperature in 5% skim milk in TBST and incubated with either of the three anti-NF antibodies (1:500) or mouse monoclonal anti-Bcl-2 (1:500; Transduction Labs, Mississauga, Ontario, Canada), purified mouse anti-Bax (2 μg/ml; PharMingen, Mississauga, Ontario, Canada), rabbit polyclonal anti-Bcl-xL (1:600), or rabbit anti-p27 (1:1,000; Santa Cruz Biotechnology, Inc., Santa Cruz, CA) overnight at 4°C. All of the primary antibodies were diluted in TBST/0.02% azide. After this incubation, membranes were then washed three times in TBST, incubated for 30 min at room temperature with either horseradish peroxidase-conjugated donkey antirabbit IgG (1:20,000 in TBST) or sheep antimouse IgG (1:10,000 in TBST; both from Amersham Pharmacia Biotech.) and then washed four times in TBST. Finally, immunopositive bands were visualized using enhanced chemiluminescence (Amersham Pharmacia Biotech.) and X-ray film (Kodak).
Immunoprecipitation and Western Blotting for Detection of FGF Receptors.
Exponentially growing cells were collected by scraping with a rubber policeman, washed free of medium with PBS, and pelletted by centrifugation. The cell pellets were then homogenized in ice-cold modified radioimmunoprecipitation assay buffer [50 mm Tris-HCl (pH 7.4), 1% NP-40, 0.25% SDS, 150 mm NaCl, and 1 mm EGTA] supplemented with the following protease inhibitors: leupeptin (0.2 μg/ml), aprotinin (5 μg/ml), and phenylmethylsulfonyl fluoride (1 mm). Protein was additionally extracted by passing cell lysates several times through 23-gauge needles, and lysates were then clarified of cellular debris and DNA by centrifugation at 3,000 rpm at 4°C for 15 min. Protein concentration in the supernatants was determined as above. Supernatants were diluted to a final protein concentration of 1.5 mg/ml in modified radioimmunoprecipitation assay buffer and aliquotted in replicates (1 ml/tube) for immunoprecipitation. To cell supernatants the following rabbit polyclonal antireceptor antibodies were added at the indicated final dilutions and allowed to incubate for 2 h at 4°C: anti-FGFR1 sc-121 (1:200; Santa Cruz Biotech.), anti-FGFR2 (1:1,000; Sigma Chemical Co.), anti-FGFR3 (1:500; Sigma Chemical Co.), or anti-FGFR4 sc-124 (1:200; Santa Cruz Biotech.). After this, 20 μl of protein A-Agarose se-2001 (Santa Cruz Biotech.) were added per tube and allowed to incubate with constant shaking overnight at 4°C. The beads were then collected and washed several times by centrifugation at 13,000 rpm with ice-cold PBS. Immunoprecipitates were eluted from the beads with sample buffer, boiled, clarified by spinning, separated by 7.5% SDS-PAGE run under reducing conditions, and then transferred electrophoretically onto Immobilon-P membranes. Membranes were then blocked by incubating in 5% skim milk/TBST for 45 min at 37°C, washed once in TBST, and then incubated overnight at 4°C with the appropriate antireceptor antibodies alone and, when commercially available, in the presence of a 5-fold excess of specific blocking peptides for the identification of nonspecific immunoreactivity. The primary antireceptor antibodies were used at the following dilutions in TBST/0.02% azide/0.15% skim milk: anti-FGFR1 (1:200), anti-FGFR2 (1:2,000), anti-FGFR3 (1:500), and anti-FGFR4 (1:400). Purified blocking peptides for FGFR1 (121P) and FGFR4 (124P) were obtained from Santa Cruz Biotech. After washing, membranes were then incubated for an additional 45 min at room temperature in horseradish peroxidase-conjugated donkey antirabbit IgG (1:20,000 in TBST), washed repeatedly in TBST, and immunoreactive bands finally visualized using enhanced chemiluminescence and autoradiography.
Tumoral Tissue Characterization and Classification.
The cerebellar tumors from which HSJ and SYR cell lines originate were both comprised primarily of small, undifferentiated cells with extremely high nuclear:cytoplasmic ratios (Fig. 1). Whereas the mitotic index was moderate to high for HSJ with on average two to five mitotic figures noted per field, it was found to be extremely low for SYR. Aborted pseudo-rosettes were evident in HSJ (Fig. 1,C) suggestive of its neuroblastic nature. Generalized cytoplasmic immunoreactivity for synaptophysin was noted in the majority of the neoplastic cells from HSJ and ∼30% of those from SYR. In contrast, GFAP immunoreactive sites colocalized only to the vascular tissue in both tumors. Furthermore, the majority of the neoplastic cells in both tumor specimens were intensely NSE immunopositive. On additional examination, two distinct histopathological patterns were clearly apparent. SYR, like our UM-MB1 studied previously (37), is characterized mostly by solid sheets of undifferentiated tumor cells (Fig. 1, A and B). Although intervening stroma and evidence of proliferative vasculature with scant reticulin fibers were occasionally noted in this tumor (Fig. 1,B), the complete absence of lucent islands strongly indicate that SYR, like UM-MB1 (Fig. 1,A), meets all of the histological criteria of the classic variant of MB. In contrast, the biphasic nature of HSJ (Fig. 1 C) as evidenced by highly cellular components containing a dense intercellular reticulin fiber network interspersed with reticulin-free nodules or “pale islands” of low cellularity throughout the majority of the tumor, indicate that HSJ represents the desmoplastic variant of MB.
Cell Lines That Are Good Models for MB.
The three cell lines we developed and characterized represent predominantly adherent cultures (HSJ > UM-MB1 > SYR), which differ in size, shape, and rates of division. Although all three of the MB lines, like their original tumors, express antigens consistent with a neuronal etiology, differences were noted in the intensity of staining for some neural-specific antigens and the expression of others (Table 1). In this regard, UM-MB1 is unique in that it does not express any MAP2, and SYR is the only cell line that is immunonegative for NF-H. Whereas vimentin and MAP5 immunoreactive sites are noted in all three of the lines, their staining intensity is greatest in UM-MB1 and HSJ, respectively. As reported previously, UM-MB1 are poorly differentiated round to oval-shaped cells, which measure 15–20 μm in diameter and extend short neurite-like processes (37). Our new MB cell line, SYR, is comprised of small (mean cell diameter = 10–12 μm), immature, round, often multinucleated cells, which do not, at the light microscopic level, appear to have any process outgrowth. In contrast with these two cell lines that originate from classic tumors, the cells in HSJ are consistently larger (mean soma diameter = 20–25 μm), more mature (evident by their lower nuclear:cytoplasmic ratios), and extend long bipolar fibrous-like processes, which measure on average 75–100 μm. Doubling times, determined from exponentially growing HSJ and SYR cultures by linear regression are 29 h (r = 0.98) and 49.8 h (r = 0.96), respectively.
FGF-2 and FGF-9 Suppress Growth of Cell Lines Derived from Classic MB Tumors.
The growth-suppressing activities of all five of the physiologically relevant FGFs were determined for the three MB cell lines after 7 days of exposure. For comparative purposes, all of the FGFs were added to the cultures at a final concentration of 0.625 nm, the concentration of bFGF/FGF-2 that produces maximal inhibition of UM-MB1 cell growth (37). The time of exposure was set constant at 7 days, the time at which growth inhibition with FGF-2 for UM-MB1 was greatest (37). In all of the instances, cells were seeded at appropriate densities to maintain logarithmic phase growth during the entire experiment and to assure that overgrowth, especially in controls, was not an issue. As clearly shown in Fig. 2, of the five FGFs tested, only two, FGF-2 and FGF-9, possess antitumoral activity for the two cell lines derived from the classic tumors, namely, UM-MB1 and SYR. In contrast, all five of the FGFs are slightly yet significantly mitogenic for HSJ, the cell line that originates from the desmoplastic MB tumor (Fig. 2).
To fully characterize mechanistically the responses of our two classic MB cell lines to FGF-2 and FGF-9, we proceeded to compare, in parallel cultures, time-dependent changes in viable cell number and cell death. The concentration of the FGFs used was set constant at 0.625 nm, one which produces maximal inhibition of the growth of both cell lines (data not shown). As shown in Fig. 3, both growth factors significantly decrease viable cell counts of UM-MB1 or SYR from day 4 or day 3 onwards, respectively. However, growth factor-induced cell death was only evident in UM-MB1 and found to be significantly greater than controls for both growth factors by day 7 (Fig. 3).
FGF-2 and FGF-9 Advance UM-MB1 Differentiation before Triggering Death.
We showed previously that FGF-2 suppresses UM-MB1 cell growth while it advances its maturation (37). We report presently that UM-MB1 responds to FGF-9 in an identical fashion (Fig. 4). As seen in the photomicrographs of viable cells (Fig. 4,A) or in the fluorescent micrographs of fixed cultures for the immunocytochemical detection of NF-M (Fig. 4,B), by day 4 of treatment, both growth factors induce comparable changes in cell morphology and density. The appearance of neurite-like extensions with growth cone-like structures and varicosities is noted by day 3 or 4 and is seen concomitantly with an increase in the levels of NF-M as determined on Western blots (Fig. 4 C).
Degenerative changes in UM-MB1 are noted late after FGF-2 or FGF-9 treatment. Evidence of cell death (Fig. 5,A) occurs in sparse growth factor-treated cultures by day 7 and reflects, as determined by the in situ detection of fragmented DNA (Fig. 5,B), an apoptotic mode of cell death. Comparative efficacy of the two FGFs in promoting apoptosis in UM-MB1 was determined by quantitating apoptotic nuclei in these cultures. The proportion of immunoreactive apoptotic nuclei is significantly higher in either FGF-2 [38.5% ± 3.2 (mean ± SE); P < 0.001; ANOVA] or FGF-9 [62.7% ± 3.4; P < 0.001; ANOVA] treated cultures when compared with untreated controls (0.17% ± 0.06) on day 7 (Fig. 5,C). In addition, FGF-9 produced significantly more apoptosis of UM-MB1 when compared with equimolar amounts of FGF-2 (P < 0.001; ANOVA, post-hoc Tukey-Kramer; Fig. 5 C).
The expression of Bcl-2, Bcl-xL, and Bax, anti- and proapoptotic products of the bcl-2 family of cell death-regulating genes were examined in our MB cell lines. Whereas all of the apoptotic effectors are visualized by Western blot analyses in UM-MB1 and SYR, Bcl-2 remains virtually undetectable in HSJ (Fig. 5,D). Consistent with the apoptotic response of UM-MB1 to the FGFs detailed above, levels of the proapoptotic Bax increase markedly in UM-MB1 when treated for 7 days with either FGF-2 or FGF-9, whereas neither Bcl-2 nor Bcl-xL expression is altered. Although quantitatively, FGF-9 is the more potent inducer of apoptosis in UM-MB1 (see Fig. 5,C), no obvious differences in expression levels of Bax were evident by this semiquantitative Western blot analysis (Fig. 5,D). Interestingly, Bcl-2 expression increases in SYR cultures after FGF-2 or FGF-9 treatment when compared with same day controls (Fig. 5,D). This increase is noted early, after 4 days of FGF treatment, and is maintained up until day 7, the last day examined (Fig. 5 D). Neither FGF-2 nor FGF-9 alters levels of Bax or Bcl-xL in HSJ nor do they induce the expression of Bcl-2 in this MB cell line.
FGF-2 and FGF-9 Inhibit SYR Proliferation without Advancing Differentiation or Inducing Cell Death.
The antitumoral effects of FGF-2 and FGF-9 on SYR as shown in FigS. 2 and 3 reflect changes in cell proliferation. This is reflected by a marked increase in the doubling time of SYR, which goes from 49.8 h (r = 0.96) in control to 57.6 h (r = 0.98) or 55.4 h (r = 0.99) in FGF-2 (0.625 nm) or FGF-9 (0.625 nm)-treated cultures, respectively. In contrast to that seen with UM-MB1, SYR does not appear to undergo any changes in cytodifferentiation when treated with FGF-2 or FGF-9 (see Fig. 6,A). The only notable change evident in these viable cultures is the marked decreases in cell densities after growth factor treatment when compared with same day controls (Fig. 6 A). Both control and FGF-treated cells remain intact, round, and viable for up to 7 days. When the experiments are extended for longer times, viability is only impacted in control cultures where overgrowth becomes an issue.
The antiproliferative activity of FGF-2 (Fig. 6,B) and FGF-9 (data not shown) are confirmed by the significant decrease in the incorporation of [3H]thymidine into FGF-treated SYR cells when compared with same day controls. These decreases are first significant by day 3 of growth factor treatment (Fig. 6,B), the same time when the first significant decreases in viable cell number are noted (see Fig. 3).
In the developing rodent cerebella an intracellular mechanism that stops cerebellar granule cell precursor proliferation and initiates their differentiation has been shown to implicate p27/Kip1 (42). In this regard, an inverse correlation has been shown to exist between cerebellar granule cell precursor proliferation and their expression levels of p27 (42). Once differentiated, induction of apoptosis of these cerebellar neurons is associated with a decrease in cellular p27 (43). In view of these findings and the cerebellar origin of this tumor, we proceeded to determine whether FGF-induced changes in proliferation in MB were accompanied with any changes in the expression of this CDK inhibitor. As shown in Fig. 6,C, in UM-MB1 cells treated with FGF-2 or FGF-9 for 4 culture days, a time when cell number is decreased yet viability unaltered, p27 levels increase significantly. At later times of growth factor treatment (day 6) just before apoptosis, p27 levels in UM-MB1 return to control values (data not shown). In SYR, by 4 days of treatment, both FGFs markedly increase p27 levels above those detected in controls (Fig. 6,C) and inhibit cell proliferation (Fig. 3). In contrast to that seen with UM-MB1, increased levels of p27 are maintained in SYR treated with the FGFs up to day 7 (last day examined; data not shown). Interestingly, and consistent with its proliferative status, untreated control or FGF-treated HSJ cells do not express immunodetectable levels of this CDK inhibitor (Fig. 6 C).
Differential Expression of FGF Receptors in MB Cells May Correlate with Responsiveness to Select Members of the FGF Family.
Four tyrosine kinase receptor types coded by four distinct genes, FGFR1–4, have been identified and shown to be involved in the FGF signaling network. Alternative splicing of the genes encoding for FGFR1, FGFR2, and FGFR3 increases the number and complexity of this family by generating variants with distinct binding and signaling capacities (44). By immunoprecipitation and Western blot analyses the expression of FGFR types in our three cell lines was investigated. Using polyclonal affinity purified antibodies against either COOH-terminal sequences of the human FGFR1 and FGFR4, or amino acid residues in the extracellular juxtamembrane portions of human FGFR2 and FGFR3, we presently show that the three cell lines do indeed exhibit differences in their expression of these receptors (Fig. 7). In this regard, whereas UM-MB1 and SYR, the cell lines that originate from classic tumors and respond selectively to FGF-2 and FGF-9 with growth inhibition express all four of the receptor types, HSJ does not (Fig. 7). Specific FGFR1 immunoreactive bands migrating at approximately Mr 125,000–130,000 and 100,000–110,000 that are detected in both UM-MB1 and SYR are not seen at all in HSJ (Fig. 7). Furthermore, although the higher MW FGFR2 immunoband is seen in all three of the cell lines at approximately Mr 130,000–140,000, the mobilities of the lower MW immunoreactive bands differ between the classic and desmoplastic tumors appearing at Mr 115,000 in UM-MB1 and SYR and as low as Mr 105,000 in HSJ (Fig. 7). One specific immunoband is detected for either FGFR3 or FGFR4. When comparing expression levels of these receptors among the cell lines, it appears that no differences were evident for FGFR4, which migrates at approximately Mr 145,000. In contrast, whereas FGFR3 is clearly present in UM-MB1 and SYR, it is barely visible in HSJ (Fig. 7). Treatment of these cells with either FGF-2 or FGF-9 does not alter expression of any of these receptors (data not shown).
We report presently that two of the five members of the FGF family found in the developing cerebellum, namely, FGF-2 and FGF-9, possess antitumoral activity for human MB. This activity is only evident for the two cell lines that express all four of the FGFR types and originate from the classic and most common variant of MB. In contrast, the MB cell line derived from a desmoplastic tumor, HSJ, is growth stimulated by these same FGFs, barely expresses FGFR3, and does not have any detectable FGFR1. Although FGF-2 and FGF-9 ultimately inhibit the progression of both classic cell lines, they only advance differentiation and promote death of one, UM-MB1. Finally, the FGFs induce changes in intracellular modulators in these tumor cells that are consistent with these responses, and support the origin of MB from the neuroprogenitors of the developing cerebellum.
Although these lines represent the clonal expansion of select tumoral cells, they are representative models for MB retaining many of the properties of most of the transformed cells in the original tumor specimens. For examples:
b.f.(1) The state of differentiation of UM-MB1 and SYR, as evidenced at the light and electron microscopic level (37) with their high nuclear:cytoplasmic ratios, sparsity of cytoplasm, rough endoplasmic reticulum and Golgi (Ref. 37; data not shown) is consistent with the immature neoplastic cells found in classic variants of this tumor (36). In contrast, the pale islands of desmoplastic MB represent areas of more advanced cellular maturity (11, 45) and HSJ, the cell line derived from this variant, exhibits a more advanced state of cytodifferentiation. However, cells of these pale islands have been documented to have low Ki-67 labeling indices (11) and to be p27 immunoreactive (46), whereas the doubling time for HSJ is the shortest of our three cell lines, and the cells are p27 immunonegative. Nevertheless, although the proliferation rates of the cell lines may appear inconsistent with their states of maturity, they are consistent with the mitotic indices of the original tumor specimens, which are highest for HSJ and lowest for SYR.
b.f.(2) In addition, the predominant immunophenotype of our three MB cell lines, like that of their original tumors, is neuronal-like, similar to that noted for the majority of MB lines generated to date (37, 47, 48, 49, 50, 51, 52, 53). Although it is difficult to speculate on the significance of the subtle differences in the antigenic profiles of our cells, the higher expression of vimentin in UM-MB1 and SYR, cells derived from classic MB, is consistent with that documented in the literature for this tumor variant (54). This high level of expression for vimentin, an intermediate filament found in neuroepithelial progenitors and neural precursors that appears before that of the NFs in cells derived from classic MBs is noteworthy, because it corroborates the proposed etiology of this variant from a more primitive germinal layer of the cerebellum, the VM (3). On the other hand, the complete absence of NF-H in SYR may hold both structural and developmental significance for a cell that is round, devoid of processes, and unable to extend neurites when exposed to the FGFs.
The expression of the FGFR subtypes in our different cell lines not only supports their etiology from distinct germinal zones in the cerebellum but may also serve to predict the type of response they will have when exposed to the FGFs. For example, FGFR transcripts are high in the VM and markedly lower in the EGL (55). This expression pattern strongly suggests that FGF may be more important in the initial stages of neuroglial differentiation of VM progeny and, thus, may preferentially target classic MB tumors. It may also explain why the classic cell lines express all four of the receptor types, whereas the desmoplastic does not. On the basis of the ligand specificities of the FGFRs, the IIIc variants of these receptors, which preferentially bind FGF-2 and -9 (56), are the isoforms that likely mediate the effects of these FGFs on UM-MB1 and SYR. In support of this, and in agreement with the histogenesis of MB from the developing cerebellum, within the central nervous system, FGFR1, FGFR2, and FGFR3 are expressed predominantly in their IIIc forms (57). The expression of the specific splice variants of these receptors in MB is currently under investigation in our laboratory using reverse transcription-PCR and splice-specific restriction digests.
Progeny of VM, like those of the subventricular zones of the neocortex, may gain competence to undergo neuronal differentiation only after they are exposed to FGF-2 (58). However, only one of the two classic MB cell lines undergo neuronal-like maturational changes when exposed to the FGFs. This selective response may once again depend on receptor expression patterns or the status of the transformed cell itself. In the case of the former, although UM-MB1 and SYR express similar immunobands for all four of the FGFRs, the lower MW FGFR1 immunoband is the more prevalent form of this receptor in UM-MB1, whereas the higher MW FGFR1 variant is more evident in SYR. These two bands may represent the 3 or 2 immunoglobulin loop variants of FGFR1, or simply different glycosylated forms of the same receptor. Although ligand binding sites are contained in immunoglobulin-2 and immunoglobulin-3 loops (59), and specificity determined by the immunoglobulin-3 loop (56), the immunoglobulin-1 loop controls binding affinity, and its removal generates a more efficient receptor with a higher affinity for its ligands (60). Therefore, should the lower MW form of FGFR1 represent a 2 immunoglobulin loop variant, it may induce more efficient downstream signaling. Its higher affinity and lower rate of dissociation may result in sustained activation of the mitogen activated protein kinases Erk1 and Erk2. In agreement with this proposal, sustained, rather than transient activation of Erk has been shown to be involved in neuronal differentiation paradigms such as neurite outgrowth (61).
The differentiation response of UM-MB1 to the FGFs may prime these cells to undergo apoptosis. In this regard, during embryonic and early postnatal periods, more than half of all neuronal precursors are eliminated by an active process called “programmed cell death” (62). MB, which is thought to originate from such neuronal precursors, may have arisen as a result of failed programmed cell death. FGFs may reinstate that response in this tumor by any one or combinations of the following mechanisms:
b.f.(1) FGF may act to initiate differentiation, a process which not only primes cell cycle exit in neuronal progenitors but increases their expression of apoptotic effectors and mediators (reviewed in Ref. 63). In support of this, we find FGF-2 or -9 to increase the levels of Bax in UM-MB1 cells committed to die. Although Bax has been shown to play essential roles in a number of different neuronal death paradigms (64, 65, 66, 67) including those that take place in developing cerebellar neurons (68), its expression must always be considered in relation to that of antiapoptotic effectors in determining cell fate (reviewed in Ref. 69). In this regard, as FGFs increase Bax in UM-MB1 without modulating Bcl-2 and Bcl-xL, the overall response presumably favors death.
b.f.(2) FGF may prime differentiation of MB yet alone be insufficient to maintain a viable fully differentiated state. In support of this, FGF has been shown to initiate neuronal-like differentiation in primary chromaffin cell cultures (70) and MAH, a sympathoadrenal progenitor cell line (71), creating a responsiveness and dependence in these cells on other neuroactive substances for their long-term survival. In the absence of these other substances, these cells, like our MB lines treated only with FGF, die.
b.f.(3) FGF may trigger apoptosis in MB tumors by promoting their differentiation while they continue to divide. Such a conflict has been thought to induce inappropriate cell cycle genes that initiate the signaling pathways leading to apoptosis (72). FGF-2 has been shown to act similarly in differentiated oligodendroglia, which make an attempt to re-enter the cell cycle but, instead, undergo apoptosis (73). In addition, although neurotrophins do not promote any evident advancement of MB differentiation, they may initiate maturation but ultimately induce apoptosis by a similar mechanism (74, 75).
The antimitotic activity of FGF for UM-MB1 and SYR may be mediated by a number of different mechanisms as well such as:
b.f.(1) FGF-2 could antagonize the activity of Sonic hedgehog, which has been found to increase the proliferation and prevent the differentiation of cerebellar granule cell precursors (76, 77).
b.f.(2) The FGFs may slow MB growth by activating an intracellular mechanism implicating p27 that normally stops cerebellar granule neuronal precursors from dividing and initiates their differentiation (42). Interestingly, induction of p27 is downstream of the TGF-β signaling pathway (78) and the growth inhibitory autocrine loop that is normally mediated by this cytokine in MB is lost in hyperdiploid tumors (79) like our own (37, 80). Therefore, it is possible that the FGFs may reinstate that autocrine loop by up-regulating release of TGF-β from MB in a manner similar to that seen with human proximal tubular cells (81). In support of this, preliminary findings from immunoneutralization studies suggest that FGF-2 may suppress UM-MB1 cell growth by increasing availability of bioactive TGF-β.4
b.f.(3) In SYR cells, FGF may slow cell proliferation by increasing expression of Bcl-2. In this regard, Bcl-2 has been shown to possess a cell cycle inhibitory role, which is distinct and independent of its role as an antiapoptotic effector (82). This antiproliferative activity has rendered Bcl-2 a good prognostic indicator for several types of tumors such as lymphomas (83) and breast cancer (84). The complete absence of Bcl-2 in HSJ may underlie its rapid doubling rate.
The cytostatic effects of exogenously added FGF have been evidenced for other tumor cells as well such as breast (66, 85, 86) and prostate (87). For breast cancer, FGFR1, the receptor that is expressed exclusively in our responsive cell lines, mediates these cytostatic responses and has been shown to be a favorable prognostic indicator for this form of cancer (88). In view of these findings, the prognostic value of this receptor for MB needs to be additionally investigated.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Supported by Fonds de la Recherche en Santé du Québec (to R. L. K.), Telethon des Étoiles (studentship to S. M. D.), and Fondation Charles-Bruneau (to Y. T.).
The abbreviations used are: MB, medulloblastoma; CDK, cyclin-dependent kinase; EGL, external granular layer; FGF, fibroblast growth factor; Ho, Hoechst; TBST, Tris-buffered saline/0.2% Tween 20; FGFR, fibroblast growth factor receptor; GFAP, glial fibrillary acidic protein; MW, molecular weight; MAP, microtubule-associated protein; NF, neurofilament; NSE, neuron-specific enolase; TGF-β, transforming growth factor β; VM, ventricular matrix.
R. L. Kenigsberg and S. M. Duplan, unpublished observations.
|Antibody specificity .||UM-MB1 .||SYR .||HSJ .|
|Antibody specificity .||UM-MB1 .||SYR .||HSJ .|
Determined on cell homogenates by Western blotting as well as by immunocytochemistry (see “Materials and Methods” for details).
We thank Liliane Gallant for typographical assistance.