Fibroblast growth factors redirect retinal axons in vitro and in vivo

Available online at www.sciencedirect.com R

Developmental Biology 263 (2003) 24 –34

www.elsevier.com/locate/ydbio

Fibroblast growth factors redirect retinal axons in vitro and in vivo C.A. Webber, M.T. Hyakutake, and S. McFarlane* Department of Cell Biology and Anatomy, Genes and Development Research Group, University of Calgary, Calgary, Alberta, Canada, T2N 4N1 Received for publication 15 January 2003, revised 1 July 2003, accepted 1 July 2003

Abstract Growth factors have been shown previously to participate in the process of axon target recognition. We showed that fibroblast growth factor receptor (FGFR) signaling is required for Xenopus laevis retinal ganglion cell (RGC) axons to recognize their major midbrain target, the optic tectum [neuron 17 (1996), 245]. Therefore, we have hypothesized that a change in expression of a fibroblast growth factor (FGF) at the entrance of the optic tectum, the border between the diencephalon and mesencephalon, may serve as a signal to RGC axons that they have reached their target. To determine whether RGC axons can sense changes in FGF levels, we asked whether they altered their behavior upon encountering an ectopic source of FGF. We found that in vivo RGC growth cones avoided FGF-misexpressing cells along their path, and that FGF-2 directly repelled RGC growth cones in an in vitro growth cone turning assay. These data support the idea that RGC axons can sense changes in FGF levels, and as such provide a mechanism by which FGFR signaling is involved in RGC axon target recognition. © 2003 Elsevier Inc. All rights reserved. Keywords: Growth cone; Xenopus; Axon guidance; Chemorepulsion; Retina; Fibroblast growth factor

Introduction The growth cone, a motile sensory apparatus at the tip of a developing axon, interprets guidance cues in its environment. Such cues attract or repel growth cones, and act either locally or at a distance from their site of production to direct the growing axon through the brain to its target. In turn, the effects of extrinsic cues on a growing axon are influenced by the intrinsic state of the growth cone (Gomez and Spitzer, 1999; McFarlane, 2000; Song et al., 1997, 1998). Growth factors are one type of extrinsic cue that guide axons. In vitro, the growth cones of Xenopus spinal cord neurons are attracted by concentration gradients of several growth factors, including brain-derived neurotrophic factor (BDNF) and neurotrophin-3 (NT-3) (Song et al., 1997). Moreover, several in vivo studies have illustrated a guidance role for growth factors, particularly in target recognition (McFarlane and Holt, 1997). For instance, in BDNF null mice, vestibular ganglion axons fail to innervate their sensory epithelial target, a normal source of BDNF (Ernfors

* Corresponding author. Fax: ⫹1-403-270-0737. E-mail address: [email protected] (S. McFarlane). 0012-1606/$ – see front matter © 2003 Elsevier Inc. All rights reserved. doi:10.1016/S0012-1606(03)00435-4

et al., 1994). Further, target recognition by sympathetic axons is dependent on nerve growth factor (NGF): Sympathetic axons genetically engineered to misexpress NGF grow out to, but fail to innervate their normal targets (Hoyle et al., 1993). In addition, hepatocyte growth factor/scatter factor (HGF/SF) in limb mesenchyme attracts incoming chick motor axons (Ebens et al., 1996). Finally, we showed previously that fibroblast growth factor receptor (FGFR) signaling is required for Xenopus laevis retinal ganglion cell (RGC) axons to recognize their major midbrain target, the optic tectum (McFarlane et al., 1996). Inhibition of FGFR signaling in RGC axons not only caused defects in their ability to extend, but also led to target recognition errors (McFarlane et al., 1996). Several pieces of evidence have led us to hypothesize that a change in fibroblast growth factor (FGF) levels at the entrance to the optic tectum (i.e., the border between the diencephalon and the mesencephalon) informs RGC axons that they have arrived at their target. First, the expression pattern of a number of FGFs changes at the entrance to the optic tectum. For instance, we found that, in Xenopus, FGF-2 protein is expressed at high levels along the optic pathway, but at much lower levels in the optic tectum (McFarlane et al., 1995). In contrast, Xenopus FGF-8

C.A. Webber et al. / Developmental Biology 263 (2003) 24 –34

mRNA is expressed at the anterior border of the optic tectum (Christen and Slack, 1997). Additionally, FGF-15 mRNA is expressed in both the embryonic mouse diencephalon and mesencephalon, but is absent from the border between the two regions (Ford-Perriss et al., 2001; McWhirter et al., 1997). Second, we found that Xenopus RGC axons fail to recognize their normal target when either growth cone FGFRs are inhibited, or when the precise spatial pattern of FGF expression is lost by broad application of FGF-2 to the diencephalon as the axons approach the optic tectum (McFarlane et al., 1995, 1996). Taken together, these data suggest that a change in FGF levels at the entrance to the optic tectum acts as a target recognition cue for RGC axons. Previous work has shown similarly that a change in NGF levels is important for murine sympathetic axons to recognize their target (Hoyle et al., 1993). To determine whether RGC axons can sense changes in FGF levels in vivo, we asked whether Xenopus RGC axons alter their behavior when encountering an ectopic source of FGF near or in the optic pathway. Guidance defects were observed as RGC axons encountered brain neuroepithelial cells that were made to misexpress FGFR ligands. Additional experiments showed that FGF-2 directly repelled RGC growth cones in vitro. These data support our hypothesis that RGC axons can sense changes in FGF levels, and as such provide a mechanism by which FGFR signaling is involved in RGC axon target recognition.

25

used both as a reporter for transgene-expressing cells and as a control. DNA transfections and electroporation

Materials and methods

Jelly coats of stage 24 embryos were removed by washing in 2% cysteine (pH 8.0), and embryos were then anesthetized in Modified Barth’s Saline [MBS; 8.8 mM NaCl, 0.1 mM KCl, 0.7 mM CaCl2, 0.1 mM MgSO4, 5 mM Hepes (pH 7.8), 25 mM NaHCO3] supplemented with 0.4 mg/ml tricaine (ethyl 3-aminobenzoic ethyl ester, methanesulfonate salt; Sigma). The DNA transfection technique has been described previously (Holt et al., 1990). In brief, DNA was mixed with a synthetic cationic lipid (DOSPER or DOTAP; Boehringer) at a 1:3 weight to volume ratio of DNA:transfection reagent. A borosilicate glass needle, pulled on an electrode puller (Sutter Instruments), was used with a Picospritzer II (General Valve Company) to make multiple extracellular injections of the DNA mixture into stage 19 or 24 embryos, targeting the neuroepithelium that gives rise to the diencephalon/midbrain. To obtain higher numbers of expressing cells, the embryo was electroporated (adapted from Haas et al., 2001) following the last injection. Briefly, a total of four to eight stimulations of 1 s each at the settings of 1 ms duration, 500 Hz and 150 V was applied to the brain by using a S44 stimulator (Grass Instruments). Embryos were allowed to develop at room temperature in 0.1⫻ MMR until stages 33/34 or 40, and were then fixed overnight in 4% paraformaldehyde in 1⫻ phosphate-buffered saline (PBS, pH 7.4) at 4°C.

Animals

Visualization of the optic projection

Embryos were attained by fertilizing eggs obtained from adult female X. laevis injected with human chorionic gonadotropin (Intervet). Embryos were kept in 0.1⫻ Marc’s Modified Ringer’s solution (MMR; 0.1 M NaCl, 2 mM KCl, 1 mM MgCl2, 5 mM Hepes, pH 7.5) with the temperature varied between 14 and 27°C to control their speed of development. Embryos were staged according to Nieuwkoop and Faber (1994).

To visualize the optic projection, RGC axons were anterogradely labeled by using horseradish peroxidase (HRP, type IV; Sigma) as described previously (Cornel and Holt, 1992). Briefly, the lens of the right eye was surgically removed, and HRP, dissolved in 1% lysolecithin (Sigma), was placed in the eye cavity. After allowing time for anterograde labeling of RGC axons (20 min), embryos were fixed overnight at 4°C in 4% paraformaldehyde in 1⫻ PBS (pH 7.4). The brains were then dissected for whole-mount immunocytochemistry.

Constructs The coding region of Xenopus embryonic fibroblast growth factor (eFGF) (kindly provided by J. Slack; Isaacs et al., 1992) was subcloned into the Stu/Xba sites of a modified CS2⫹ cDNA expression vector (Turner and Weintraub, 1994) to generate CS2-eFGF. CS2⫹ plasmids containing either the coding regions of Xenopus fibroblast growth factor-2 (FGF-2; kindly provided by J. Slack) (Patel and McFarlane, 2000; Slack et al., 1987) or Xenopus FGFR ligand-2 (FRL-2; kindly provided by M. Kirschner) (Kinoshita et al., 1995) were also used. A cDNA construct encoding green fluorescent protein (GFP; CS2-GFP) was

BrdU labeling Transfected embryos were injected in the gut with 5 mg/ml bromodeoxyuridine (BrdU; Sigma) diluted in MBS, at stage 33/34. Two hours following injection, embryos were fixed in 4% paraformaldehyde in 1⫻ PBS overnight at 4°C. Twenty-micrometer transverse cryostat sections were treated for 20 min with 2 M HCl at room temperature and rinsed before being processed for immunocytochemistry with an anti-BrdU antibody (Amersham Pharmacia Biotech).

26

C.A. Webber et al. / Developmental Biology 263 (2003) 24 –34

C.A. Webber et al. / Developmental Biology 263 (2003) 24 –34

27

Fig. 2. GFP-positive cells cotransfected with the FGF-2 cDNA plasmid reliably misexpress the FGF-2 transgene. (A–B) Lateral views of whole-mount stage 40 Xenopus brains transfected with GFP (A) or FGF-2 (B) plasmid at stage 24. HRP-labeled optic projections ignore GFP-misexpressing cells (arrowhead in A), but avoid an FGF-2-misexpressing cell (arrowheads in B). (C–D) Transverse sections through the eyes and brain of a stage 33/34 embryo transfected at stage 19 with GFP and FGF-2 cDNA plasmids. The majority of GFP-positive cells (C) are also immunolabeled with an antibody that recognizes Xenopus FGF-2-misexpressing cells (D). Tec, tectum; ot, optic tract; Di, diencephalon; R, retina; D, dorsal; A, anterior; V, ventral. Scale bar in (A) is 50 ␮m for (A–B). Scale bar in (C) is 25 ␮m for (C–D).

Immunocytochemistry Immunocytochemistry was performed either on wholemount brains or on 12- to 20-␮m cryostat transverse sec-

tions through the embryonic diencephalon and midbrain. Embryos for cryostat sectioning were rinsed twice in 1⫻ PBS (pH 7.4) and cryoprotected for 30 min in 30% sucrose. Prior to sectioning, embryos were frozen in optimal cutting

Fig. 1. RGC axons avoid FGF-misexpressing cells in vivo. Whole-mount stage 40 Xenopus embryo brains shown from a lateral view. Representative examples of the behavior of HRP-labeled RGC axons (red) encountering either GFP (A–B), or GFP and eFGF (C–F) plasmid-expressing brain cells (green). (B) and (D) are higher power views of boxed areas in (A) and (C), respectively. Brain cells transfected with both GFP and eFGF plasmids appear to have differentiated and exhibit processes (arrows). Tec, tectum; ot, optic tract; Tel, telencephalon; Di, diencephalon; A, anterior; D, dorsal. Scale bar in (A) is 50 ␮m for (A, C, and E). Scale bar in (B) is 25 ␮m for (B, D, and F).

28

C.A. Webber et al. / Developmental Biology 263 (2003) 24 –34

Fig. 3. A non-FGF FGFR ligand also redirects RGC axons. (A–E) Lateral views of whole-mount Xenopus brains transfected with GFP (A), or GFP and FRL-2 (B–E) cDNA plasmids at stage 19. HRP-labeled RGC axons (red) ignored neuroepithelial cells misexpressing GFP (A, green), but avoided crossing over the GFP and FRL-2-misexpressing cells (B and C, green). (C) shows the overlay of photomicrographs of the HRP-labeled optic tract (D) and the GFP/FRL2misexpressing cell (E). (F) Graph showing the percentage of optic projections that deviated around transfected cells misexpressing GFP alone (n ⫽ 14), GFP and eFGF (n ⫽ 11), FGF-2 (n ⫽ 10), and GFP and FRL-2 (n ⫽ 8). Only cells misexpressing an FGFR ligand affected the RGC optic projection. D, dorsal; A, anterior; ot, optic tract. Scale bar is 25 ␮m for (A, C–E). Scale bar in (B) is 50 ␮m.

C.A. Webber et al. / Developmental Biology 263 (2003) 24 –34

29

Adobe Photoshop (4.0) software. Primary antibodies used in this study were as follows: mouse monoclonal anti-HRP (Sigma) at a dilution of 1:2000 (0.22 mg/ml), mouse monoclonal anti-BrdU (Amersham Pharmacia Biotech) at a dilution of 1:4, and rabbit polyclonal FGF-2 antibody (R&D Systems) at a dilution of 1:100 (10 ␮g/ml). All antibodies were visualized with the fluorescent mouse anti-Rhodamine Red X (RRX) secondary antibody at a dilution of 1:500 (3 ␮g/ml) (Jackson ImmunoResearch Laboratories, Inc.). Retinal cultures Eye primordia were dissected from stage 24 Xenopus embryos and plated as explant tissue or dissociated cells on poly-L-ornithine (Sigma) coverslips coated with 10 ␮g/ml laminin (Sigma) for neurite measurements (as per Harris et al., 1985), or the cells were grown on 50 ␮g/ml fibronectin (Roche) for growth cone turning assays. For explant cultures, culture media consisted of 60% L-15-glutamine (Invitrogen), 0.1% BSA, and 1% antibiotic–antimycotic (Invitrogen). For neurite outgrowth assays, dissociated retinal cultures were grown either in the above solution (control) or in control solution supplemented with either 20 or 250 ng/ml human recombinant FGF-2 (Invitrogen). After 24 h, cultures were fixed for 45 min in 1% gluteraldehyde and mounted in aquamount (Baxter). Analysis was performed blind and involved measuring the longest uninterrupted neurite of solitary RGCs using Spot II software (Diagnostic Instruments). RGCs were easily identified on the basis of morphology (Worley and Holt, 1996). Samples were statistically analyzed by using an ANOVA test, followed by a Student-Newmann-Keuls post-hoc test. Fig. 4. FGF-misexpressing cells are postmitotic. Twenty micrometer transverse sections through the eyes and midbrain of Xenopus embryos transfected at stage 19 with GFP, or GFP and eFGF plasmids (green), and injected with BrdU at stage 33/34. After 2 h, stage 35/36 embryos were fixed, sectioned, and processed for BrdU immunocytochemistry (red). Few neuroepithelial cells misexpressing GFP (A), or GFP and eFGF (B), were BrdU-positive. Mb, midbrain; R, retina; D, dorsal; V, ventral; Ve, ventricle. Scale bar is 50 ␮m for (A–B).

temperature (OCT) compound (Baxter), and quick-frozen at ⫺20°C. Standard immunostaining procedures were used for cryostat sections and whole-mount brains (Cornel and Holt, 1992; McFarlane et al., 1995). Samples were incubated overnight at 4°C in primary antibody diluted in PBT [PBS, 0.1% bovine serum albumin (BSA) (Sigma), 0.5% Triton (BDH)] with 5% goat serum (Invitrogen). Fluorescent secondary antibodies were applied for 1 h at room temperature. After washing in PBT, samples were mounted in glycerol with an anti-bleaching agent, p-phenylenediamine [1 mg/ml in 9 parts glycerol, 1 part 1 M Tris–HCl (pH 8.5); Sigma]. Digital images of samples were obtained by using a Spot II camera and Spot Advanced software (Diagnostics Instruments), and processed for brightness and contrast by using

Growth cone turning assay Stage 24 retinal explant cultures were used in the growth cone turning assay 18 –24 h after plating (as per de la Torre et al., 1997). Growth cones that actively grew in a straight line for 30 min before the beginning of the experiment were chosen. The responses of these growth cones to an applied concentration gradient were recorded for 45 min by using a Cohu CCD video camera and Scion Image capture software (shareware). Stable FGF-2 gradients were formed by pulsatile ejection of 0.1 mg/ml recombinant human FGF-2 protein (Invitrogen) from a 0.5- to 1-␮m tip glass capillary pipette placed at a 45° angle from the actively extending growth cone (Lohof et al., 1992; Zheng et al., 1994). The pressure was applied with an electrically gated pressure application system (Picospritzer General Valve). A standard pressure pulse of 3 p.s.i. in amplitude was applied for 20 ms to the pipette at a frequency of 2 Hz using a pulse generator (SD9, Grass Instruments). By this method, stable concentration gradients can be established from the micropipette, and, at a distance of 50 –100 ␮m away, the growth cone sees a concentration approximately 1000-fold lower than in the

30

C.A. Webber et al. / Developmental Biology 263 (2003) 24 –34

Fig. 5. Low and high FGF-2 concentrations stimulate RGC neurite extension in vitro. Stage 24 eyebuds were dissociated and cultured for 24 h in 60% L-15-glutamine with 0.1% BSA, or in this media supplemented with 20 or 250 ng/ml human recombinant FGF-2. RGCs were easily identified as having a large, phase bright cell body with one to three long processes (as per Worley and Holt, 1996). The length of their longest uninterrupted neurite was measured. Lengths averaged for five to six independent experiments are shown. Both low and high concentrations of FGF-2 significantly stimulated neurite extension over and above that observed in control (*, P ⬍ 0.05, ANOVA, Student-Newmann-Keuls post-hoc test).

pipette (i.e., 100 ng/ml) (Lohof et al., 1992; Zheng et al., 1994). Control solutions consisted of 1% BSA in 1⫻ PBS. Upon completion of the experiment, only actively extending growth cones (⬎5 ␮m growth as per Zheng et al., 1994) were analyzed. Experiments were performed blind so that only after the analysis of the recorded video was the identity of the pipette solution revealed to the experimenter. For analysis, the trajectories of the growth cones were traced onto a graph and the turning angles were measured.

Results RGC axons avoid FGF-misexpressing cells in vivo Based on our previous FGF signaling gain-and loss-offunction data (McFarlane et al., 1995, 1996), we hypothesize that a change in FGF levels at the anterior border of the midbrain acts as a target recognition signal for incoming RGC axons. To determine whether RGC axons can sense changes in FGF levels in vivo, ectopic sources of an FGF were provided along the optic pathway. At stage 19 or 24, we transfected brain neuroepithelial cells with a cDNA plasmid encoding the secreted FGF protein Xenopus embryonic-FGF (eFGF) (Isaacs et al., 1992). eFGF is not expressed in the optic pathway, but given that FGFRs are promiscuous with respect to ligand binding, ectopically produced eFGF should activate axonal FGFRs (Greene et al., 1998). While it is likely that other misexpressed growth factors could affect the behavior of RGC axons, for example BDNF (Cohen-Cory and Fraser, 1994), the purpose of our experiments was to test whether this is the case for FGFs. To visualize transgene-expressing cells, a reporter plasmid

encoding green fluorescent protein (GFP) was cotransfected into the neuroepithelial cells. Transfection of the GFP plasmid alone served as the control. At stage 40, when normally the optic tract has innervated the target, RGC axons were anterogradely labeled with HRP, and whole-mount brains were processed with an antibody that recognizes HRP. In controls, RGC axons that encountered GFP-positive cells in the optic pathway ignored the cells and grew along their normal path toward the optic tectum (Figs. 1A and B, 2A, and 3A; n ⫽ 14 embryos). In contrast, RGC axons consistently turned to avoid eFGF-misexpressing cells (Fig. 1C–F; n ⫽ 11 embryos). It was clear in some cases that the axons turned at a distance of at least one cell diameter from the eFGF-expressing cell (Fig. 1D and E). Axons were misdirected whether they encountered single (Fig. 1C–E) or multiple (Fig. 1F) eFGF-misexpressing cells, and whether the cells were in the optic tract (Fig. 1C and D) or optic tectum (Fig. 1E and F). These data indicate that an ectopic FGF source affects RGC growth cone behavior at a distance. To confirm that the GFP-positive cells also misexpressed FGF protein, we performed identical experiments with a construct encoding Xenopus FGF-2 (Kimelman and Kirschner, 1987; Slack et al., 1987), as an antibody that recognizes this protein is commercially available (Patel and McFarlane, 2000). FGF-2-overexpressing cells caused defects in the trajectories of RGC axons similar to those observed with eFGF misexpression; however, the axons came into closer proximity to the FGF-2-expressing cells before they altered their trajectory (Fig. 2B; n ⫽ 10 embryos). Immunocytochemistry revealed that 93% of GFP-positive cells coexpressed the FGF-2 protein (n ⫽ 200 cells; Fig. 2C and D). Given this high coexpression rate, and the similar avoidance phenotype observed with all the eFGF and FGF-2 but none of the GFP-transfected embryos, it is likely that a comparable coexpression rate is obtained with eFGF and GFP coinjections. Misexpression of a non-FGF ligand that interacts with FGFR also redirects RGC axons Several proteins that are not members of the FGF family can also bind to and activate FGFRs. One of these is the FGFR ligand-2 (FRL-2) (Kinoshita et al., 1995). To determine whether ectopic expression of FRL-2 could also misdirect RGC axons, we cotransfected FRL-2 and GFP cDNA plasmids into the developing Xenopus neuroepithelium at stages 19 or 24 and assayed the behavior of HRP-labeled axons at stage 40. The cells misexpressing the reporter protein GFP did not affect RGC axon trajectories (Fig. 3A; n ⫽ 14). In contrast, in all cases of RGC axons encountering ectopic FRL-2-expressing cells, the axons preferred to grow on adjacent cellular substrate (Fig. 3B–E; n ⫽ 8 embryos). While we were unable to confirm that the cells were expressing FRL-2, we feel confident that this was the case, since avoidance was observed whenever axons encountered

Fig. 6. FGFs directly repel RGC growth cones in vitro. RGC growth cones extending from a 24-h, stage 24 explant culture. A pipette with a 0.5- to 2.0-␮m opening (black star) was used to eject in a pulsatile fashion either control media (A–B), or an FGF-2-containing media (C–D), onto an extending RGC growth cone. (A, C) Growth cones prior to applying the pipette solution. (B, D) Growth cones 45 min after continuous exposure to the applied pipette solution. The growth cone does not respond to the control solution (A–B), but is redirected away from the pipette ejecting FGF-2 (C–D). (E–F) Superimposed neurite trajectories of growth cones exposed to control (E) and FGF-2 (F)-containing pipettes. Scale bar in (A) is 25 ␮m.

32

C.A. Webber et al. / Developmental Biology 263 (2003) 24 –34

FRL-2-expressing cells but never when encountering GFPexpressing cells, and the FGF-2 construct showed high coexpression rates with GFP (Fig. 2C and D). These data indicate that different FGFR ligand families can redirect RGC axons in vivo (Fig. 3F).

FGF-2 concentration that failed to promote rat cerebellar neurite outgrowth was still stimulatory to RGC axon extension (Williams et al., 1994). Therefore, it is unlikely that the FGF-misexpressing cells affected the development of the optic tract by inhibiting axon extension.

The FGF-misexpressing neuroepithelial cells are postmitotic

FGFs directly alter RGC growth cone trajectories in vitro

There are several explanations for the misdirection of RGC axons caused by FGF-misexpressing cells. First, FGFs are known to act as strong mitogens in the developing nervous system (Gospodarowicz and Bialeki, 1979; Michler-Stuke and Bottenstein, 1982). It is possible that FGF-misexpressing cells and their neighbors remain mitotically active, so that the diencephalic environment would be unlike that of the stage 33/34 brain through which the RGC axons normally extend. To test this possibility, at stage 33/34, we injected eFGF and/or GFP-transfected embryos with bromodeoxyuridine (BrdU), which labels proliferative cells. The GFP-expressing cells (Fig. 4A) and the GFP/ eFGF coexpressing cells (Fig. 4B) showed little colabeling with BrdU (12%, n ⫽ 128; 7%, n ⫽ 135, respectively), indicating that in both cases the vast majority of transfected cells were postmitotic. Indeed, many of the GFP/eFGF coexpressing cells had a differentiated morphology, with long processes (Fig. 1D and F, arrows). These data argue against the possibility that the FGF-misexpressing cells remained undifferentiated and thus failed to provide the appropriate cues for RGC axon pathfinding. High and low FGF concentrations stimulate RGC axon extension in vitro FGF-2 is known to stimulate neurite extension of cultured Xenopus RGCs and rat cerebellar neurons (McFarlane et al., 1995; Williams et al., 1994); however, at high doses, FGF-2 fails to promote extension in rat cerebellar neuron cultures (Williams et al., 1994). In our in vivo experiments, FGFs were misexpressed at high levels. As such, it is possible that the ectopic FGF could be inhibitory to RGC axon extension and thus cause defects in the development of the optic tract. To determine whether high concentrations of FGFs adversely affect RGC axon extension, we cultured dissociated stage 24 retinal cultures in the presence of low (20 ng/ml; n ⫽ 6 independent cultures) or high (250 ng/ml; n ⫽ 5 independent cultures) levels of human recombinant FGF-2 protein. We showed previously that the former concentration promotes RGC neurite outgrowth (McFarlane et al., 1995). RGCs are easily identifiable in the dissociated cultures (Worley and Holt, 1996), and the length of their longest neurite was measured. There was no obvious difference in neurite length when the cultures were exposed to high or low concentrations of FGF-2 (Fig. 5; *, P ⬍ 0.05, ANOVA, Student-Newmann-Keuls post-hoc test). Thus, an

To determine whether FGFs act as a chemotropic molecule for RGC axons, we used an in vitro growth cone turning assay (Lohof et al., 1992; Zheng et al., 1994). Eyebuds were explanted at stage 24 and cultured for 18 –24 h before performing the turning assay on RGC growth cones that had extended from the explant (de la Torre et al., 1997). Sister embryos allowed to develop for the same time period would be at stage 33/34, and their RGC axons would be extending through the diencephalon (Holt, 1989). Previously, we showed that cultured RGC growth cones express FGFRs (McFarlane et al., 1996). RGC growth cones were first recorded for 30 min by using time-lapse videomicroscopy. Subsequently, either control media (0.1% BSA in PBS) or control media supplemented with 0.1 mg/ml FGF-2 was applied to one side of the growth cone by pulsatile ejection, and growth cones were recorded for an additional 45 min. In the majority of cases (87.5%, n ⫽ 8), control media did not affect the growth cone trajectory (Fig. 6A, B, and E). In contrast, several RGC growth cones (50%, n ⫽ 8) turned away from the FGF-2-containing pipette (Fig. 6C, D, and F). A trace of the paths taken by all growth cones in response to a control or an FGF-2 source is represented in Fig. 6E and F. The mean turning angle of RGC growth cones when presented with an FGF-2 source (⫺17.9° ⫾ 8.1° s.e.m, n ⫽ 8) was significantly different from that of control axons (1.0° ⫾ 1.7° s.e.m, n ⫽ 8; P ⬍ 0.05, unpaired, two-tailed Student’s t test). These data indicate that FGFs directly repel RGC axons in vitro, and suggest that FGFdependent chemorepulsion is the explanation for why RGC axons avoid FGF-misexpressing cells in vivo.

Discussion In this study, we demonstrated that developing RGC axons sense changes in FGF levels in vivo and in vitro. We showed that in vivo, FGF-misexpressing cells along the optic tract or within the optic tectum caused RGC axon pathfinding errors, presumably in an FGFR-dependent manner. In vitro, we illustrated that FGFs directly repel RGC growth cones in a growth cone turning assay. These data indicate that RGC axons are sensitive to FGF levels both in vivo and in vitro and that FGFs can act as chemorepellents. In vivo, the ectopic FGF could influence RGC axons directly or indirectly through a signal produced by the transgene-expressing cell or its neighbors. For the following reasons, we favor the former explanation. First, we found

C.A. Webber et al. / Developmental Biology 263 (2003) 24 –34

that RGC growth cones turned away from an FGF-2 source in vitro. Importantly, previous work has shown that, under similar experimental conditions, RGCs cultured on fibronectin grew toward, not away from, a source of the growth factor netrin (Hopker et al., 1999), arguing for the specificity of the FGF-2 effect. This is the first demonstration that FGFs are chemotropic for developing axons, though FGF-10 has been shown to act as a chemoattractant for distal epithelial buds during lung development (Park et al., 1998). Second, the BrdU labeling experiments suggest that the differentiation of FGF-misexpressing cells and their neighbors was unaffected, despite the known mitogenic activity of FGFs (Gospodarowicz and Bialeki, 1979; Michler-Stuke and Bottenstein, 1982). As such, it seems likely that the avoidance of FGF-misexpressing cells by RGC axons in vivo is also explained by a direct chemorepulsive action of FGF on RGC axons. The fact that the RGC axons often appeared to change their direction at a distance from eFGF-expressing cells supports this idea. However, the FGF-2- and FRL-2-misexpressing cells did not affect axon behavior at a distance, leaving the possibility that in vivo, a different mechanism, such as contact-mediated repulsion, is occurring. The fact that members of two different molecular families, FGFs and FRL-2, produced the same repulsion of RGC axons implicates a signal transduction cascade initiated through FGFR–ligand interaction. The demonstration that RGC growth cones extending in vitro, and presumably in vivo, can sense changes in local FGF concentration lends further support to our model that it is a change in FGF levels at the border to the optic tectum that serves as a target recognition cue for RGC axons. Indeed, there is a change in the expression of several FGFs, including FGF-2, FGF-8, and FGF-15, at the border between the diencephalon and mesencephalon (Christen and Slack, 1997; McFarlane et al., 1995; McWhirter et al., 1997). Given that target recognition defects are observed when the pattern of FGF expression is disrupted (McFarlane et al., 1995), this ability to sense a change in FGF concentration may be critical for normal RGC axon target recognition. How might FGF chemotropism be involved in RGC axon target recognition? While we found that exogenous FGFs were chemorepulsive in vivo and in vitro, it does not necessarily mean that the endogenous FGFs at the border to the optic tectum similarly act as chemorepellents. Growth cone behavior in response to specific cues appears to be context-dependent. For instance, altering intracellular growth cone cyclic nucleotide levels dramatically affected how cultured Xenopus growth cones responded to gradients of netrin and BDNF (de la Torre et al., 1997; Song et al., 1997). In both cases, lowering cyclic adenosine monophosphate (cAMP) converted an attractive turning response to one of repulsion. In addition, the response of cultured RGC growth cones to netrin depended on the levels of the extracellular matrix molecule laminin (Hopker et al., 1999).

33

Thus, FGFs could act as either an attractant or a repellent for RGC axons reaching the anterior border of the optic tectum. Previous studies suggest explanations for why we found in vitro that FGFs applied locally in a concentration gradient were repulsive to RGC axons and yet FGF-2 induced RGC neurite outgrowth when bath applied at a uniform concentration. First, the internal state of the growth cone could affect one process but not the other. For instance, lowering cAMP activity changed netrin from an attractive to a repulsive influence on Xenopus spinal cord neuron growth cones, but did not affect the rate of growth cone extension (Ming et al., 1997). Second, different FGFR intracellular signaling cascades may participate in RGC axon extension and turning. Indeed, the phosophinositol-3-kinase (PI3K) pathway is important for FGF-dependent Xenopus RGC axon extension but not for RGC target recognition (Lom et al., 1998). In addition, glial cell line-derived neurotrophic factor promotes survival and differentiation of pheochromocytoma cells through the mitogen-associated protein kinase pathway and the PI3K pathway, respectively (Chen et al., 2001). Finally, it is possible that an axon’s response to a cue depends on how that cue is presented to the cell or growth cone. For instance, FGF-2 was chronically bath applied to the entire cell in the extension assay, but was applied transiently to one side of the growth cone in the turning assay. In summary, this study is the first to show that FGFs affect the trajectories of axons extending both in vitro and in vivo, and that FGF can act as a chemorepellent. Classically, in the absence of cyclic nucleotide manipulation, growth factors such as BDNF and NT-3 are considered chemoattractant molecules (Song et al., 1997). In the future, it will be important to determine whether FGFs act chemotropically to influence the target recognition of RGC axons. Acknowledgments We thank Dr. C.C. Logan, J.C. Hocking, A.C. Behie, N.S. Pollock, R.D. Parker, and C.L. Hehr for helpful comments on the manuscript and Dr. C.E. Holt for intellectual input. C.L. Hehr and A.C. Behie provided valuable technical assistance. We thank Dr. J. Slack for the eFGF and FGF-2 plasmids, and Dr. M. Kirschner for the FRL-2 plasmid. C.A.W. was supported by an Alberta Heritage Foundation for Medical Research (AHFMR) studentship, and the Alberta Odd Fellow Rebekah Visual Research Fund, and S.M. by Canadian Institutes of Health Research (CIHR) and AHFMR scholarships. This work was supported by a CIHR operating grant and by funding from the Lion’s Sight Centre. References Chen, Z., Chai, Y., Cao, L., Huang, A., Cui, R., Lu, C., He, C., 2001. Glial cell line-derived neurotrophic factor promotes survival and induces

34

C.A. Webber et al. / Developmental Biology 263 (2003) 24 –34

differentiation through the phosphatidylinositol 3-kinase and mitogenactivated protein kinase pathway respectively in PC12 cells. Neuroscience 104, 593–598. Christen, B., Slack, J.M., 1997. FGF-8 is associated with anteroposterior patterning and limb regeneration in Xenopus. Dev Biol 192, 455– 466. Cohen-Cory, S., Fraser, S.E., 1994. BDNF in the development of the visual system of Xenopus. Neuron 12, 747–761. Cornel, E., Holt, C., 1992. Precocious pathfinding: retinal axons can navigate in an axonless brain. Neuron 9, 1001–1011. de la Torre, J.R., Hopker, V.H., Ming, G.L., Poo, M.M., Tessier-Lavigne, M., Hemmati-Hemmati-Brivanlou, A., Holt, C.E., 1997. Turning of retinal growth cones in a netrin-1 gradient mediated by the netrin receptor DCC. Neuron 19, 1211–1224. Ebens, A., Brose, K., Leonardo, E.D., Hanson Jr., M.G., Bladt, F., Birchmeier, C., Barres, B.A., Tessier-Lavigne, M., 1996. Hepatocyte growth factor/scatter factor is an axonal chemoattractant and a neurotrophic factor for spinal motor neurons. Neuron 17, 1157–1172. Ernfors, P., Lee, K.F., Jaenisch, R., 1994. Mice lacking brain-derived neurotrophic factor develop with sensory deficits. Nature 368, 147– 150. Ford-Perriss, M., Abud, H., Murphy, M., 2001. Fibroblast growth factors in the developing central nervous system. Clin. Exp. Pharmacol. Physiol. 28, 493–503. Gospodarowicz, D., Bialeki, H., 1979. Fibroblast and epidermal growth factors are mitogenic agents for cultured granulosa cells of rodent, porcine, and human origin. Endocrinology 104, 757–764. Gomez, T.M., Spitzer, N.C., 1999. In vivo regulation of axon extension and pathfinding by growth-cone calcium transients. Nature 397, 350 –355. Greene, J.M., Li, Y.L., Yourey, P.A., Gruber, J., Carter, K.C., Shell, B.K., Dillon, P.A., Florence, C., Duan, D.R., Blunt, A., 1998. Identification and characterization of a novel member of the fibroblast growth factor family. Eur. J. Neurosci. 10, 1911–1925. Haas, K., Sin, W.C., Javaherian, A., Li, Z., Cline, H.T., 2001. Single-cell electroporation for gene transfer in vivo. Neuron 29, 583–591. Harris, W.A., Holt, C.E., Smith, T.A., Gallenson, N., 1985. Growth cones of developing retinal cells in vivo, on culture surfaces, and in collagen matrices. J. Neurosci. Res. 13, 101–122. Holt, C.E., 1989. A single-cell analysis of early retinal ganglion cell differentiation in Xenopus: from soma to axon tip. J. Neurosci. 9, 3123–3145. Holt, C.E., Garlick, N., Cornel, E., 1990. Lipofection of cDNAs in the embryonic vertebrate central nervous system. Neuron 4, 203–214. Hopker, V.H., Shewan, D., Tessier-Lavigne, M., Poo, M., Holt, C., 1999. Growth-cone attraction to netrin-1 is converted to repulsion by laminin-1. Nature 401, 69 –73. Hoyle, G.W., Mercer, E.H., Palmiter, R.D., Brinster, R.L., 1993. Expression of NGF in sympathetic neurons leads to excessive axon outgrowth from ganglia but decreased terminal innervation within tissues. Neuron 10, 1019 –1034. Isaacs, H.V., Tannahill, D., Slack, J.M., 1992. Expression of a novel FGF in the Xenopus embryo. A new candidate inducing factor for mesoderm formation and anteroposterior specification. Development 114, 711– 720. Kimelman, D., Kirschner, M., 1987. Synergistic induction of mesoderm by FGF and TGF-beta and the identification of an mRNA coding for FGF in the early Xenopus embryo. Cell 51, 869 – 877.

Kinoshita, N., Minshull, J., Kirschner, M.W., 1995. The identification of two novel ligands of the FGF receptor by a yeast screening method and their activity in Xenopus development. Cell 83, 621– 630. Lohof, A.M., Quillan, M., Dan, Y., Poo, M.M., 1992. Asymmetric modulation of cytosolic cAMP activity induces growth cone turning. J. Neurosci. 12, 1253–1261. Lom, B., Hopker, V., McFarlane, S., Bixby, J.L., Holt, C.E., 1998. Fibroblast growth factor receptor signaling in Xenopus retinal axon extension. J. Neurobiol. 37, 633– 641. McFarlane, S., 2000. Attraction vs. repulsion: the growth cone decides. Biochem. Cell Biol. 78, 563–568. McFarlane, S., Cornel, E., Amaya, E., Holt, C.E., 1996. Inhibition of FGF receptor activity in retinal ganglion cell axons causes errors in target recognition. Neuron 17, 245–254. McFarlane, S., Holt, C., 1997. Growth factors: a role in guiding axons? Trends Cell Biol. 7, 424 – 430. McFarlane, S., McNeill, L., Holt, C.E., 1995. FGF signaling and target recognition in the developing Xenopus visual system. Neuron 15, 1017–1028. McWhirter, J.R., Goulding, M., Weiner, J.A., Chun, J., Murre, C., 1997. A novel fibroblast growth factor gene expressed in the developing nervous system is a downstream target of the chimeric homeodomain oncoprotein E2A-Pbx1. Development 124, 3221–3232. Michler-Stuke, A., Bottenstein, J.E., 1982. Proliferation of glial-derived cells in defined media. J. Neurosci. Res. 7, 215–228. Ming, G.L., Song, H.J., Berninger, B., Holt, C.E., Tessier-Lavigne, M., Poo, M.M., 1997. cAMP-dependent growth cone guidance by netrin-1. Neuron 19, 1225–1235. Nieuwkoop, P.D., Faber, J., 1994. Normal Table of Xenopus laevis. Garland Publishing Inc., New York. Park, W.Y., Miranda, B., Lebeche, D., Hashimoto, G., Cardoso, W.V., 1998. FGF-10 is a chemotactic factor for distal epithelial buds during lung development. Dev. Biol. 201, 125–134. Patel, A., McFarlane, S., 2000. Overexpression of FGF-2 alters cell fate specification in the developing retina of Xenopus laevis. Dev. Biol. 222, 170 –180. Slack, J.M., Darlington, B.G., Heath, J.K., Godsave, S.F., 1987. Mesoderm induction in early Xenopus embryos by heparin-binding growth factors. Nature 326, 197–200. Song, H., Ming, G., He, Z., Lehmann, M., McKerracher, L., TessierLavigne, M., Poo, M., 1998. Conversion of neuronal growth cone responses from repulsion to attraction by cyclic nucleotides. Science 281, 1515–1518. Song, H.J., Ming, G.L., Poo, M.M., 1997. cAMP-induced switching in turning direction of nerve growth cones. Nature 388, 275–279. Turner, D.L., Weintraub, H., 1994. Expression of achaete-scute homolog 3 in Xenopus embryos converts ectodermal cells to a neural fate. Genes Dev. 8, 1434 –1447. Williams, E.J., Furness, J., Walsh, F.S., Doherty, P., 1994. Characterisation of the second messenger pathway underlying neurite outgrowth stimulated by FGF. Development 120, 1685–1693. Worley, T., Holt, C., 1996. Inhibition of protein tyrosine kinases impairs axon extension in the embryonic optic tract. J. Neurosci. 16, 2294 – 2306. Zheng, J.Q., Felder, M., Connor, J.A., Poo, M.M., 1994. Turning of nerve growth cones induced by neurotransmitters. Nature 368, 140 –144.