Alternate FGF2-ERK1/2 signaling pathways in retinal photoreceptor and glial cells in vitro.

Basic fibroblast growth factor (FGF2) stimulates photoreceptor survival in vivo and in vitro, but the molecular signaling mechanism(s) involved are unknown. Immunohistochemical and immunoblotting analyses of pure photoreceptors, inner retinal neurons, and Müller glial cells (MGC) in vitro revealed differential expression of the high affinity FGF receptors (FGFR1-4), as well as many cytoplasmic signaling intermediates known to mediate the extracellular signal-regulated kinase (ERK1/2) pathway. FGF2-induced tyrosine phosphorylation in vitro exhibited distinct profiles for each culture type, and FGF2-induced ERK1/2 activation was observed for all three preparations. Whereas U0126, a specific inhibitor of ERK kinase (MEK), completely abolished FGF2-induced ERK1/2 tyrosine phosphorylation and survival in cultured photoreceptors, persistent ERK1/2 phosphorylation was observed in cultured inner retinal cells and MGC. Furthermore U0126 treatment entirely blocked nerve growth factor-induced ERK1/2 activation in MGC, as well as FGF2-induced ERK1/2 activation in cerebral glial cells. Taken together, these data indicate that FGF2-induced ERK1/2 activation is entirely mediated by MEK within photoreceptors, which is responsible for FGF2-stimulated photoreceptor survival. In contrast, inner retina/glia possess alternative, cell type, and growth factor-specific MEK-independent ERK1/2 activation pathways. Hence signaling and biological effects elicited by FGF2 within retina are mediated by cell type-specific pathways.

Basic fibroblast growth factor (FGF2) 1 belongs to a family of structurally related polypeptides, encompassing at present over 20 factors (1,2) that stimulate growth and differentiation of cells of mesodermal and neuroectodermal origin (3). In the central nervous system, FGF2 is expressed widely in neuronal and glial cells (4,5) and possesses neuroprotective effects both in vivo (6) and in vitro (7). FGF2 mediates its biological effects via binding and activation of specific high affinity tyrosine kinase receptors, named FGFR1-4 (8,9).
FGF signal transduction has been intensively studied, mostly in non-neuronal tissue, using immortalized cell lines overexpressing one or more FGFR (10 -12), and a variety of molecules involved in the FGF signal transduction cascade have been described: phospholipase C␥1 (PLC␥1), a regulator of phosphatidyl inositol mechanism (13); the adaptor protein Shc; son of sevenless (SOS), a guanine nucleotide exchange factor for ras and growth factor receptor-binding protein 2 (Grb2) (12); a phosphotyrosine phosphatase designated as SH-PTP2 (14); and protein kinase B (Akt) (15). These molecules activate downstream signaling pathways, including those of the mitogen-activated protein kinases (16), also known as extracellular signal-regulated kinases (ERK), a family of serine/ threonine protein kinases. The ERK signal transduction cascade is common to many different cell types including neurons (17)(18)(19). Mammalian ERK1 (p44) and ERK2 (p42) are the best known and best studied members of the MAPK family (20). Activation of ERK1 and ERK2 occurs via phosphorylation by dual-specificity MAPK kinases, MEK1 and MEK2, and is often associated with the stimulation of cell proliferation, differentiation, and survival (21)(22)(23).
Within the central nervous system, especially the retina, FGF2 has shown promise for therapeutic treatment of neurodegeneration. FGF2 delays photoreceptor (PR) breakdown in various rat models in vivo (24 -26), and PR-targeted FGFR inactivation leads to retinal degeneration in mice (27). In vitro, FGF2 induces PR differentiation (28) and directly stimulates survival of purified PR (29). The complexity of the central nervous system has rendered the analysis of molecular pathways underlying neurotrophic actions very difficult, but their understanding is essential for the development of rational therapeutic strategies. Although recent studies have implicated ERK activation in FGF2 signaling (30,31), for the moment there is no evidence that PR survival is directly influenced via the ERK pathway. We have exploited primary cultures of different retinal cells (purified PR, inner retina (IR, without PR), and purified Mü ller glial cells (MGC)) to determine whether FGF2-related signaling molecules have distinct expression patterns between the different cell populations and whether FGF2 possesses multiple pathways for signal transduction in the retina. This is indeed the case, PR survival depending uniquely upon ERK activation via MEK. Whereas in FGF2-treated PR, MEK seems to be the only upstream ERK activator, additional MEK-independent pathways exist in IR and MGC.
Animals-Wistar rats used for the experiments were cared for and handled in compliance with the Association for Research in Vision and Ophthalmology statement for the Use of Animals in Ophthalmic and Vision Research. Animals were deeply anesthetized by CO 2 inhalation and killed by cervical dislocation.
Immunohistochemistry-Cultures of PR and IR were fixed after 3 days in vitro for 15 min with 4% paraformaldehyde. For immunostaining cultured cells were pemeabilized with Triton X-100 (0.1% in PBS for 5 min) and then saturated with PBS containing 1% BSA, 0.1% Tween 20, and 0.1% NaN 3 (blocking solution). Samples were incubated overnight at 4°C with the following primary antibodies diluted in blocking solution to a final concentration of 10 g/ml: anti-FGFR1, anti-FGFR2, anti-FGFR3, anti-FGFR4, anti-SOS1, anti-SOS2, anti PLC␥1, anti-SH-PTP2, anti-Shc, anti-ERK1/2, anti-Akt, and anti-Grb2. For controls, primary antibodies were preadsorbed with their corresponding immunizing peptides used at 50-fold higher concentration. Primary antibodies were localized using goat anti-mouse IgG or goat anti-rabbit IgG/Alexa Fluor TM 488 secondary antibodies. Immunostained slides and cells were examined using a Nikon Optiphot 2 photomicroscope equipped with fluorescence and Nomarski optics.
Isolation and Culture of Photoreceptors and Inner Retinal Cells-PR and IR were isolated as described previously, using a modified technique (32,33) originally developed for PR transplantation (34) to obtain purified fractions of PR and IR. Briefly, PN5 retina were flat-mounted PR-down on a gelatin block and the entire preparation cooled with DMEM/ϪCO 2 . A first cut was made at a depth of 100 m from the vitreal surface to isolate the IR, containing the ganglion cell, inner plexiform, and inner nuclear layers (GCL, IPL, and INL, respectively). Fractions were collected in Ringer's solution without Ca 2ϩ supplemented with 2.5 mM EDTA and kept at 4°C. A second cut of about 50 m eliminated remaining INL, outer plexiform layer, and a part of the PR layer. A final cut of 250 -300 m undercut the PR layer still attached to the gelatin block. Fractions were incubated in Ringer's solution without Ca 2ϩ , plus EDTA at 37°C for 10 min to separate PR from the gelatin. After three washes in Ringer's solution PR and IR were digested with 500 l of activated papain (0.1 mg/ml Ringer's solution) for 20 min at 37°C. Digestion was stopped with 500 l of DMEM supplemented with 10% fetal bovine serum (DMEM/10% FBS), 50 l of DNase 1 (1 mg/ml), and 0.1% BSA and the cells incubated for 5 min at 37°C. After dissociation the cells were centrifuged at 800 rpm for 10 min and the pellet resuspended in 1 ml of serum-supplemented DMEM. Viable cell numbers were determined and the cells seeded at 5 ϫ 10 5 cells/cm 2 into poly-D-lysine-coated (2 g/cm 2 during 1 h) culture dishes. For immunohistochemistry of PR and IR and for PR survival assays, cells were seeded at lower density (10 5 cells/cm 2 ) on glass coverslips coated with poly-D-lysine (2 g/cm 2 during 1 h) and laminin (1 g/cm 2 overnight).
Culture of Glial Cells-Cultures of MGC were established using a previously described technique (35) and used after one passage (ϳ7-10 days in vitro). Cultures of mixed brain glia (astrocytes and oligodendrocytes) were prepared from neonatal rat brain cerebral cortex using previously published methods (36) and used after one passage.
Growth Factor or Inhibitor Treatment-PR and IR were cultured for 24 h in DMEM/10% FBS. Then cells were rinsed with DMEM and replaced in a chemically defined medium (CDM) (29) for 48 h. Confluent MGC and brain glial cell cultures grown in DMEM/10% FBS were rinsed with DMEM and replaced in CDM for 48 h. Cells were stimulated with FGF2 (100 ng/ml) for 30 s to 30 min. For some experiments cultures were treated with the MEK inhibitor U0126 (10 -50 M) for 30 min prior to growth factor stimulation. Two additional trials using U0126 were performed: MGC were preincubated with U0126 (10 M) for 30 min and stimulated with NGF (100 ng/ml) for 30 min, and mixed neonatal brain glia were preincubated with U0126 (10 M) for 30 min and stimulated with FGF2 (100 ng/ml) for 30 min.
Photoreceptor Survival Assay-The survival assay for PR was conducted as described previously (29). Briefly, after 24 h DMEM/10% FBS was replaced by CDM supplemented or not with FGF2 (20 ng/ml) and U0126 (100 nM to 1 M). FGF2 and U0126 were added again after 72 h. After 5 days in vitro the viability of PR was tested by the Live/Dead assay (37). For each coverslip 25 fields were recorded (using 20ϫ objective for observation) using Visiolab 1000 image analysis software (Biocom, Lyon, France), and cells were counted. For each treatment in each experiment two coverslips were counted and the experiment conducted three times. Statistical analysis was performed using the parametric Peritz' F test according to Harper (38), values of p Ͻ 0.05 being considered statistically significant.
Protein Extraction and Western Blotting-For immunodetection of FGFR1-4, Akt, ERK1/2, Grb2, MEK1, MEK2, PLC␥1, Shc, SH-PTP2, SOS1, SOS2, arrestin, and vimentin, cultures of PR, IR (3 d in vitro), and MGC were rinsed with PBS and collected in lysis buffer (20 mM Tris-HCl, pH 7.4, 150 mM NaCl, 1% Nonidet P-40, 1 mM EDTA, 1 mM NaF, 1 mM Na 3 VO 4 ) containing a protease inhibitor mixture. For antiphosphotyrosine and anti-phospho-ERK1/2 immunoblots, cultures were washed once with PBS and stimulated with growth factors for different times. The reaction was stopped by addition of liquid nitrogen. Cells were then collected in lysis buffer and lysed as above. Proteins were separated by SDS-polyacrylamide gel electrophoresis and transferred to nitrocellulose membranes. For probing with anti-phosphotyrosine antibody, the membranes were blocked with PBS, 0.2% Tween 20, 3% BSA, 1% fat-free milk powder for 1 h at room temperature. For all other antibodies membranes were blocked PBS, 0.1% Tween 20, 5% fat-free milk powder for 1 h at room temperature. Membranes were then incubated with primary antibodies (anti-FGFR, signaling intermediates, and cell type-specific markers, each 1 g/ml final concentration) overnight at 4°C. Membranes were then incubated with goat anti-mouse or goat anti-rabbit IgG-horseradish peroxidase secondary antibodies (0.15 g/ml). Immunoreactive bands were visualized using a Pierce Super Signal West Pico kit according to the manufacturer's instructions. In some trials Western blots probed with anti-phospho-ERK1/2 antibody were stripped and subsequently reprobed with anti-ERK1/2 antibody. Molecular weights were compared with prestained molecular weight markers.

FGFR and FGF-related Signaling Molecules Are Widely Expressed in Retinal Cells in
Vitro-Antisera against FGFR1-4 were used to investigate expression and distribution of FGFR and second messengers in the different culture models. FGFR1 and FGFR4 were clearly and uniformly expressed at the surface of cultured PN5 PR, whereas FGFR2 and especially FGFR3 stained only weakly (Fig. 1, A-H). Within cultures of IR, which contain a mixed population of neurons (bipolar, amacrine, and ganglion cells) and MGC, all four FGFR were detected (Figs. 1, I-P). FGFR1 stained neurons and glia with equal intensity, while FGFR2-4 labeled neurons more intensely than MGC. If antibodies were used that had been preadsorbed with their corresponding immunizing peptides, staining was completely abolished (data not shown). Immunostaining of cultured retinal cells with antisera directed against second messenger molecules (PLC␥1, SOS1, SOS2, ERK, SH-PTP2, SHC) showed uniform PR, neuronal, and glial labeling (data not shown).
Expression of FGFR1-4 and the various signaling molecules were further compared between different culture models using Western blotting techniques. Culture purity was checked using cell type-specific antibodies; arrestin immunoreactivity was specifically expressed in PR but was absent from IR and MGC cultures, whereas vimentin immunoreactivity was detected in IR and MGC but not in PR cultures ( Fig. 2A).
FGFR1 was detected in PR at a molecular mass of ϳ120 kDa. In IR, immunopositive bands were detected at 140 and 100 kDa and in MGC at 150, 125, and 100 kDa. FGFR2-positive bands were found in PR, IR, and MGC at 90 kDa. Additional bands were observed in MGC cultures at 130 and 100 kDa. Western blots probed with the FGFR4 antibody revealed a single band at ϳ120 kDa in PR, IR, and MGC. For all receptors MGC exhibited stronger signals compared with PR and IR (Fig. 2B). If antibodies were preadsorbed with their corresponding immunizing peptides prior to immunoblotting, no positive bands were observed (data not shown). We were not able to obtain satisfactory results on Western blots using the FGFR3 antibody.
Western blots for FGF-related signaling molecules revealed that despite loading of equal total protein concentrations between fractions, PR preparations showed consistently lower relative amounts of each protein examined compared with IR and MGC (Fig. 3). Levels of PLC␥1, SOS2, Akt, and Grb2 were markedly lower in PR than in IR and MGC, while levels of SOS1 and SH-PTP2 were equivalent. Specific differences in expression of ERK1/2 and Shc were observed; while ERK2 was present at similar levels between the three samples, ERK1 was notably reduced in PR; and while three isoforms of Shc at 46, 52, and 66 kDa were observed in both IR and MGC, the 46-kDa isoform was very weak and the 66-kDa isoform undetectable in PR (Fig. 3). SOS1 and Akt were more abundant in pure MGC than in mixed IR cultures.
FGF2 Induces Cell-specific Signal Cascades in Different Populations of Retinal Cells in Vitro-To examine FGF2-induced tyrosine phosphorylation, PR, IR, and MGC cultures were incubated for increasing times with a fixed dose of FGF2 and samples immunoblotted with a phosphotyrosine-specific antibody. In PR cultures, as described previously (29), we observed time-dependent tyrosine phosphorylation of major bands at ϳ140, 120, 105, 95, 74, and 65 kDa and minor bands with molecular masses of ϳ180 and 155 kDa (Fig. 4A). The most prominent bands at 140 and 120 kDa had already increased in intensity at 30 s and reached their maximum at 2 min. In IR cultures, FGF2 induced time-dependent tyrosine phosphorylation of bands at ϳ180, 120, 90, 65 and a closely spaced doublet of 55 and 53 kDa (Fig. 4B). The major band at 120 kDa had already reached maximal intensity by 30 s, remaining strongly stimulated at 2 min and decreasing by 5 min. The band at 180 kDa was phosphorylated at approximately similar levels throughout the times examined, the proteins at ϳ90 and 65 kDa demonstrated highest staining at 2 min, and the doublet at 55 and 53 kDa became maximally stimulated after 5 min. MGC cultures treated with FGF2 showed the same pattern of tyrosine-phosphorylated bands as IR and one additional band with a molecular mass Ͼ200 kDa (Fig. 4C). The band at 120 kDa showed maximal stimulation already after 30 s and maintained this level up to 5 min. The band at 65 kDa showed increased phosphorylation after 30 s and reached its maximum at 1 min. The 90-kDa band showed only a weak increase in phosphorylation between 30 s and 2 min but showed a strong phosphorylation after 5 min. The bands at 180 and Ͼ200 kDa increased steadily in intensity over the duration of incubation. The different phosphotyrosine profiles for each culture model are schematized in Fig. 6D to facilitate comparison.
FGF2 Induces Activation of ERK1/2 in Vitro-As the ERK pathway represents a major pathway for FGF2-induced signaling in neurons (19) and glia (39), we investigated whether exogenous application of FGF2 could induce phosphorylation of ERK1 and ERK2 in PR, IR, and MGC cultures in vitro. Cultures were incubated for 15 or 30 min with FGF2, and ERK1/2 activation was monitored by immunoblotting using phospho-ERK1/2-specific antibody. As shown in Fig. 5, after 15 min of stimulation IR and MGC cultures showed an increase in phosphorylated forms of ERK1/2, whereas no ERK activation was observed in PR. Within IR samples the increase was almost completely restricted to ERK2, while MGC showed equally intense labeling for both isoforms. After 30-min stimulation the phosphorylation of ERK1/2 had further increased in IR and MGC, and for IR samples was now also elevated in ERK1. Phospho-ERK1/2 was also visible in PR at this time point, predominantly in ERK2, showing the direct activation of the ERK pathway by FGF2 in these cells.

FGF2-induced ERK1/2 Activation Is Blocked Completely by the MEK Inhibitor U0126 in PR but Not in IR and MGC-Since
exogenous application of FGF2 led to activation of the ERK pathway in PR, IR, and MGC, we examined whether we could suppress the ERK1/2 activation with the MEK1-and MEK2specific inhibitor U0126 (40). Cultures of PR, IR, and MGC were preincubated with U0126 and then stimulated with FGF2 for 30 min and phospho-ERK1/2 detected as above. U0126 completely blocked ERK1/2 activation in PR at concentrations of 10 and 20 M (Fig. 6A). In contrast, in IR and MGC the phosphorylation of ERK1/2 was decreased but not completely blocked (Fig. 6A). U0126 used at 20 M led to a 2-fold decrease in ERK1/2 phosphorylation, but compared with control cultures, the activation of ERK1/2 was still marked. Even a concentration of 50 M U0126 did not further decrease ERK1/2 phosphorylation in IR and MGC (data not shown). Use of an- other MEK inhibitor PD098059 (50 and 100 M) to block ERK activation in MGC confirmed these results (data not shown).
To verify that the persistent ERK phosphorylation in IR and MGC was not due to incomplete inhibition of MEK, two different controls were performed. Neonatal rat brain glia were treated with FGF2 in the presence or absence of U0126; the inhibitor entirely blocked FGF2-induced ERK1/2 phosphorylation, with only residual background activation remaining (Fig.  6B). In the second, MGC cultures were stimulated with NGF in the presence or absence of U0126; whereas NGF induced sustained ERK1/2 activation in MGC, in cultures preincubated with U0126 this effect was completely inhibited (Fig. 6C). These two results showed clearly that U0126, at the concentrations used, efficiently blocked MEK and inhibited downstream ERK phosphorylation. No differences in protein expression levels of MEK1 and MEK2 were observed in the retinal cultures; Western blots revealed distinct and uniformly expressed bands at 42 and 45 kDa for MEK1 and MEK2, respectively, in PR, IR, and MGC (Fig. 7).
Survival-promoting Activity of FGF2 Is Mediated by the ERK Pathway in PR in Vitro-We previously reported that survival of purified PR in vitro is directly stimulated by FGF2 (29). To see whether this effect of FGF2 was mediated by ERK1/2 activation, survival of control and FGF2-treated PR after 1 or 5 days in vitro was monitored using a cytotoxicity assay kit. Survival at 5 days was chosen because the previous study showed the survival-promoting effect of FGF2 to be maximal at this time. At this time point, significantly more PR survived in the presence of 20 ng/ml FGF2 than untreated PR: 60% in the presence of FGF2 compared with 40% in controls or rescue of 50% PR that would have died in the absence of exogenous factor. If PR were cultivated in the presence of different concentrations of the MEK inhibitor U0126 (0.1-1 M), the survival-promoting effect of FGF2 was completely abolished, and the survival rate dropped to the level of untreated controls (Fig. 8).

FGF2-induced Tyrosine Phosphorylation in Retinal Cells in
Vitro-Despite the presence of all FGFR and candidate signaling intermediates in PR, IR, and MGC, FGF2-induced tyrosine phosphorylation profiles and kinetics differed between them. IR and pure MGC had largely comparable patterns of tyrosine phosphorylation, which is normal given that IR contain significant numbers of MGC. There were distinct differences between IR and MGC, however, especially the presence of prominent Ͼ200and 65-kDa phosphoproteins in pure MGC. Such changes may be due to different culture conditions (prolonged exposure to serum and increased proliferation in MGC) or lack of neuron-glia interactions. PR showed a very different pattern with several phosphorylated bands between 74 and 155 kDa, absent from IR and MGC. These various phosphotyrosine proteins could represent differentially expressed FGFR isoforms or signaling molecules, differential recruitment of common signaling molecules, or differential phosphatase activity. Our efforts to assign specific phosphotyrosine-immunoreactive bands to individual candidate proteins were not successful. Particularly the identity of the major phosphotyrosine protein at ϳ140 kDa in PR is puzzling, since this band potentially represents an FGFR through the rapidity and intensity of FGF-induced activation, yet none of the data obtained for FGFR1, FGFR2, or FGFR4 reveal the presence of such a molecular mass. A second major phosphotyrosine band of ϳ120 kDa within PR cultures  5. FGF2 induced ERK1/2 activation in vitro. Cultures from PR, IR, and MGC were stimulated for 15 or 30 min with FGF2 (100 ng/ml), and total protein extracts (20 g/lane) were analyzed by Western blot using anti-phospho-ERK1/2 antibody. After 15 min, whereas no phosphorylation of ERK1/2 could be seen in PR, IR and MGC showed an increase of phosphorylated ERK1/2 compared with the nonstimulated control. After 30-min ERK1/2 phosphorylation in IR and MGC had further increased, and ERK activation was now also detectable in PR. Western blot panels are representative of three independent experiments that gave similar results. also exhibits rapid activation kinetics and may correspond to the Western blotting data for FGFR1 and/or FGFR4. The major phosphotyrosine band in MGC at ϳ125 kDa coincides with both FGFR1-and FGFR4-immunoreactive proteins. In Western blots of FGF-treated IR and MGC, phosphotyrosine-immunoreactive bands at ϳ90 kDa match up with the FGFR2-immunopositive label. Western blots of PR demonstrated FGFR2immunoreactive bands of the same weight, but we did not observe any FGF-induced activation. The Western blotting data revealed that MGC expressed higher levels of all FGFR, as well as several signaling intermediates, than either PR or IR. Since IR cultures themselves contain many MGC, this upregulation may result from either more prolonged culture conditions or alterations in neuron-glial interactions. In summary to this section, the data demonstrate that even in highly purified and simplified neuronal and glial populations it is very difficult to attribute individual phosphotyrosine proteins to specific identified signaling molecules by such methods. Even so, the numerous differences in phosphorylation patterns and kinetics between cell types indicate the existence of distinct signaling pathways within retinal tissue.

FIG. 6. Effect of U0126 on FGF2 induced ERK activation in vitro.
A, cultures of PR, IR, and MGC were preincubated with the MEK inhibitor U0126 for 30 min and then stimulated with FGF2 (100 ng/ml) for 30 min. Total proteins (20 g/lane) were analyzed by Western blot using anti-phospho-ERK antibody. ERK activation was completely blocked in PR in the presence of 10 M U0126. In IR and MGC, phosphorylated ERK1/2 was still clearly detectable. U0126 at 20 M led to a 2-fold reduction in intensity of ERK immunolabeling in IR and MGC compared with cultures stimulated with FGF2 alone, but ERK activation was still marked. In each case (PR, IR, MGC), the upper row shows the treatments used (FGF2, U0126 at 10 or 20 M), the middle row shows immunodetection of phospho-ERK (pERK1/2), and the lower row shows immunodetection of total ERK1/2 (Western blots were stripped and re-probed with antibodies to ERK1/2). To test the efficacy of U0126 at the concentrations used, neonatal rat cortical mixed glia were stimulated with FGF2 (B), or MGC were stimulated with NGF (C) in the presence or absence of the U0126 (10 M). In both cases growth factor-induced activation of ERK1/2 was blocked totally by the MEK inhibitor. The upper row shows treatments and concentrations used (FGF2, NGF, U0126 at 10 M), the middle row shows immunodetection of pERK1/2, and the lower row shows immunodetection of total ERK1/2 (Western blots were stripped and re-probed with antibodies to ERK1/2). Panels are representative of three independent experiments that gave similar results. Statistical analysis: asterisks above FGF2-treated culture with respect to untreated control: ***, p Ͻ 0.005. Small circles above FGF2/U1026treated cells with respect to FGF2 treatment alone: three small circles, p Ͻ 0.005; two small circles, p Ͻ 0.01; one small circle, p Ͻ 0.05. ERK Signaling in Retinal Cells in Vitro-ERK1/2 are implicated in a diverse array of cellular functions, such as cell growth and proliferation, differentiation, and apoptosis (41), and activation of the ERK cascade occurs in response to different environmental stimuli, including growth factors (42,43) and neurotrophins (44,45). FGF2 application led to time-dependent ERK1/2 phosphorylation in PR, IR, and MGC in vitro. It has been reported previously that FGF2 can induce ERK activation in glia (46,47) and cortical neurons (48)

in vitro.
Other groups have demonstrated that FGF2-induced ERK activation regulates neurite outgrowth or neuronal survival in embryonic chicken retina in vitro (30,49). It has not been shown previously that FGF2 can directly induce ERK activation in PR. Indeed Wahlin et al. (31) were unable to show FGF2-induced activation of intracellular signaling pathways in PR in vivo and concluded that FGF2 exerts its effects by acting indirectly through the activation of MGC or other IR cells. FGF2 clearly induced ERK phosphorylation in purified postmitotic PR in vitro, but with slower kinetics than that observed for other retinal cells. Pretreatment of PR with the specific MEK inhibitor U0126 abolished FGF2-induced ERK1/2 phosphorylation and completely inhibited FGF2-induced PR survival, demonstrating that FGF2-dependent PR survival is mediated entirely by ERK1/2 activation via MEK.
Interestingly, U0126 pretreatment of IR and MGC reduced, but did not abolish, FGF2-induced ERK1/2 phosphorylation in IR and MGC. This was surprising, since ERK1/2 activation has been reported to be uniquely regulated by MEK (50, 51). The efficacy of U0126 as a potent inhibitor of MEK is well established (52,40), blocking ERK activation at significantly lower doses compared with another MEK inhibitor PD098059 (53). Studies with either PD098059 or U0126 have shown that MEK inhibition fully prevents FGF2-induced proliferation of neuronal progenitor cells (54) and ERK1/2 activation in mammary fibroblasts (55) and cortical neurons (48) at concentrations comparable with or even lower than those used here. The persistent ERK1/2 phosphorylation in IR and MGC was not due to incomplete inhibition of MEK, since both FGF2-induced ERK1/2 phosphorylation in brain glial cells, and NGF-induced ERK1/2 phosphorylation in MGC, were completely blocked in the presence of 10 M U0126. We chose brain glial cells as controls, since FGF2 has been shown to induce activation of ERK1/2 in astrocytes (46), and FGF2-induced proliferation of oligodendrocyte progenitors can be blocked by a MEK inhibitor (39). The persistence of FGF2-induced ERK1/2 activation within MGC even in the presence of 50 M U0126 strongly suggests FGF2 activation of ERK1/2 can operate via a MEKindependent pathway and suggests that this is related to both the cellular origin (retina versus brain) and growth factor (FGF2 versus NGF) in question. Since IR cultures also contain MGC, we do not know if the MEK-independent ERK1/2 phosphorylation in IR is only due to MGC or also to inner retinal neurons. The existence of MEK-independent pathways in ERK activation has been reported in Swiss 3T3 cells (56), myoblastic cells (57), and in a mouse hepatoma cell line (58), but to the best of our knowledge this is the first report of their presence in normal central nervous system cells. It should be noted that PD098059 fully inhibited FGF2-induced ERK activation in mixed cultures of embryonic chicken retina (59,30). This could either be due to species-specific differences in FGF2-induced ERK signaling pathways or to the appearance of MEK-independent pathways later in development. How MEK-independent activation of ERK occurs is not yet known, but may involve currently unidentified ERK-activating kinases or cross-talk between signaling pathways (60). ERK1/2 has been shown to phosphorylate upstream kinases of the ERK cascade such as Raf (61) or MEK (62) or cytoplasmic kinases such as MAPKAP kinase-2 and MAPKAP kinase-3 (63,64). In addition the presence of ERK consensus phosphorylation sites in Smad1 has been shown, connecting the ERK and Smad signaling pathways (65).
Conclusions and Perspectives-Well characterized retinal cell cultures provide unique models to analyze FGF2 signaling cascades and biological effects in the central nervous system. We have shown the presence of cell type-specific soluble messengers, and distinct tyrosine phosphorylation profiles, and pin-pointed the importance of MEK-activated ERK activity in FGF2-induced PR survival in vitro. Furthermore we provide evidence that FGF2-induced MEK-independent signal transduction pathways for ERK activation exist in the retina. Future work will focus on the identification of MEK-independent ERK activators in the retina and on the role of individual signaling molecules in PR survival through the use of dominant-negative (27) or homologous recombination technology. In addition, FGF2 plausibly acts by preventing apoptotic PR cell death, since serum deprivation induces apoptosis in primary neurons in vitro (66,67), neurotrophic factor withdrawal induces apoptosis during development (68), and FGF2 prevents apoptosis in embryonic chicken retinal cultures (30). Future studies will address how FGF2 affects pro-and anti-apoptotic pathways. Such studies will provide impetus to the application of such neurotrophic factors as potential therapies for inherited retinal degeneration.