Spatially discrete FGF-mediated signalling directs glial morphogenesis

Axons signal glia in a number of ways, inducing glial mitogenesis, differentiation and eventual morphological patterning including glial processes wrapping around axons. A number of candidate signals have been proposed which might mediate these signalling events, including FGF (Barres et al., 1993; Wang et al., 1998). FGF signalling during axonal-glial patterning has been especially complicated to study in vertebrates as this signalling pathway is used in early stages of development as well as later at multiple stages of glial differentiation (McKinnon et al., 1993; Rossant et al., 1997). However, FGF signalling is likely to be simpler in the developing insect nervous system as exemplified by the effects of a null mutation of the glial-expressed heartless FGF receptor in fruit flies (Gisselbrecht et al., 1996; Shishido et al., 1997). In addition, mutations in the one other known fly FGF receptor, breathless, affect migration of midline cells, a specialized subset of CNS glia (Klambt et al., 1992). Insect glia perform many of the physiological roles performed by vertebrate glia although there does not appear to be a clear PNS/CNS distinction between the glial types (Lane, 1981). Insect glia surround neuronal cell bodies, provide a blood brain barrier and ensheath axons (Hoyle, 1986). While insect glia do not show the same morphological complexity as vertebrate myelin, axonal glia in insects do wrap around axons (Hoyle, 1986; Lane, 1981). However, myelin-like glial structures have been identified in arthropods (Davis et al., 1999). In the fly FGF receptor mutant background, heartless, glia migrate to the major axonal pathways, distribute themselves along these fascicles, but fail to elongate and form their normal mature flat morphology (Gisselbrecht et al., 1996; Shishido et al., 1997). This indicates that in the absence of FGF signalling, glia can find axons but fail to enwrap them. Thus, FGF is a candidate signalling pathway for inducing axonal ensheathment. The data presented here supports the idea that an FGF-like signal spatially patterns glia such that they direct adhesion towards axons. Glia that line and enwrap insect axon bundles are normally elongated and flattened. In culture, addition of FGF2 to the medium, which activates the endogenous HEARTLESS FGF receptor, causes glia to rapidly round up and loose their characteristic flattened morphology. In addition, endogenous MAPK activation is localized to regions at the axonal/glial interface. Beads coated with FGF2 also induce rounding of adjacent glial cells. Based on these experiments it is proposed that it is the spatial asymmetry of FGF signalling that is critical for the morphological interaction between axons and glia.

Grasshopper embryos were obtained from a crowded colony maintained in the lab as described by Condron and Zinn (1997) except that vermiculite was used instead of sand for egg collection and well-washed store-bought lettuce used instead of lab grown wheat. Embryos were staged as described by Bentley et al. (1979) and dissected and cultured in Schneider’s insect medium (Sigma S-0146) containing 1/100 diluted penicillin/streptomycin stock (Sigma, P-0906) (Condron and Zinn, 1997). Dissected ventral nerve cords from 55% embryos were used throughout, except for the experiment described in Fig. 3A-B, which were 40%.

1 μl of 25 μm polystyrene beads (Polysciences, 7313) together with 1 μl human recombinant FGF2 (heparin stabilized; Sigma F-9786) were placed on a nutator for 60 minutes at room temperature. The beads were centrifuged and washed with 500 μl Schneider’s insect medium (Sigma S-0146) three times. A small number of beads were injected with a pipette into the medium over the dissected CNS ganglia. Beads sank very slowly and so could be spatially manipulated with turbulence currents from a passing microdissection needle as they dropped. For the experiments described in Fig. 3B beads were pushed into the CNS. Ten beads were thus placed on each dissected CNS. For the experiment in (Fig. 3C-F), beads were stuck on the ends of electrodes by stroking the surface of a bead with an air-filled, end-blocked electrode. Beads stuck to electrodes, presumably via static charges, for variable amounts of time but generally longer than the time required to maneuver them into place with a Narashige micromanipulator. Once the beads settled into place, the embryos were left still for 5 minutes at room temperature before being placed at 33°C, which is the optimum temperature for grasshopper development. Haemocyte (macrophages, as identified by non-RK2 staining and morphology) also stick to FGF2-coated beads. However, these are peripherally derived as very few were found when isolated ganglia were used in these experiments. Of 128 attached beads tested for glial staining with anti-rk2, 106 had adhered glia, 8 had non-glia adhered and 14 remained attached to the CNS with no visible cells adhered.

DIC and fluorescent images were taken with a Photometrics SenSys camera and an Olympus BX40 microscope. All images were taken with an Olympus 40× lens in conjunction with a World Precision Instruments stage warmer set to 33°C. ImageIP software was used to capture the images and fluorescent images were subjected to one frame deconvolution (95% removal, 75% gain) using VayTek Hazebuster. No further image quality processing was performed. Fluorescent channels were stacked in Adobe Photoshop and layouts/labeling performed in PowerPoint.

For each lane of western analysis (Fig. 1B), a single 55% dissected ventral nerve cord containing segments T1-A4 was placed in 100 μl lysis buffer (20 mM Tris pH 7.5, 137 mM NaCl, 1% NP-40, 5 mM

EDTA, 10% glycerol, 2 mM Na₃VO₄, 10 mM NaF, 10 μg/ml aprotonin, 20 μg/ml leupeptin, 1 μg/ml pepstatin, 1 mM PMSF) and agitated with a microtube pestle for 1 minute and then centrifuged at 14,000 rpm for 20 minutes in a microcentrifuge. The resulting supernatant was added to 200 μl PBT (1× PBS, 1 mg/ml BSA, 0.1% Triton X-100) and 1 μl polyclonal supernatant directed against the extracellular portion of the fly HTL protein. After incubation overnight at 4°C, 10 μl of protein A agarose (Sigma 1134515) was added and incubated for 2 hours at room temperature. The sample was finally microcentrifuged at 6000 rpm for 3 minutes, and washed three times with PBS. To the final sample, 30 μl of sample buffer was added (62.5 mM Tris pH 6.8, 2% SDS, 25% glycerol, 0.01% bromophenol blue) and heated at 95°C for 10 minutes, microcentrifuged for 3 minutes at 6000 rpm and then resolved by 7% PAGE. 5 μl of each sample was used for the anti-HTL lanes and 25 μl used for anti-tyrosinephosphate lanes in (Fig. 1B). The gel was electroblotted onto nitrocellulose and probed in 1 ml PBT with either 1 μl polyclonal serum directed against the intracellular portion of the fly HTL protein or 10 μl of monoclonal antibody directed against phosphotyrosine (Upstate Biotechnology 05-321) overnight at 4°C. The membrane was then washed with PBT and probed again with 1 μl of alkaline phosphatase-linked goat anti-mouse or rabbit (Jackson Labs) and incubated for 2 hours at room temperature. After washing to remove excess secondary antibody, the nitrocellulose membrane was stained using BCIP/NBT (BioRad). Stained blots were scanned into Adobe Photoshop and the average intensity of each band measured.

For immunohistochemistry, tissues were processed as described by Condron et al. (1994). All ganglia were fixed with 4% fresh formaldehyde except for anti-heartless stains which used 2% formaldehyde. It was not possible to obtain successful common fixation conditions for staining both anti-HTL and anti-activated MAPK. The rk2 (repo) glial marker was a kind gift from A. Tomlinson and was used at 1/300; anti-protein kinase A, catalytic subunit was used at 1/1000 and was a kind gift from D. Kalderon; anti MAPK-phosphate was from Sigma (M-8159), and was used at 1/900; anti-HTL antisera were a kind gift from A. Michelson and S. Gisselbrecht and were used at 1/6000; secondary antibodies were purchased from Jackson Labs and were used at 1/300.

Two FGF receptors have been cloned so far in the fruit fly (Gisselbrecht et al., 1996; Klambt et al., 1992). These are breathless (btl) and heartless (htl). In the CNS, the btl gene product (BTL) is expressed in and required for midline glial migration (Klambt et al., 1992) while htl is expressed and required for lateral glial morphogenesis (Gisselbrecht et al., 1996; Shishido et al., 1997). To examine the expression of htl in the grasshopper embryo, antiserum against the extracellular portion of the fly protein was used to stain gel-resolved protein extracts of grasshopper embryos as well as intact ganglia. A single major protein species stains with the antiserum (Fig. 1A) and in addition, the amount of phosphotyrosine is increased upon treatment of ganglia with FGF2 (data not shown). Immunoprecipitation of CNS protein extracts with the antiserum yields a single protein species (Fig. 1B, lower lanes) that also stains for phosphotyrosine (Fig. 1B, upper lanes). While the protein extract was immunoprecipitated with antiserum directed against the extracellular portion of the fly HTL protein, antiserum directed against the intracellular portion of fly HTL was used to probe the blotted protein. Protein extract from a single ventral nerve cord was used in each lane, however, samples were incubated with 38 ng/ml FGF2 for various lengths of time. The amount of phosphotyrosine (Fig. 1B upper lane) clearly increases with time, indicating that purified human FGF2 (bFGF) can activate grasshopper HTL protein. The ratio of the quantity of signal in the anti-tyrosine phosphate band over anti-HTL band is shown below the respective lanes. The smearing seen in the lower lanes is the result of contamination from anti-HTL antibody, which runs just ahead of the HTL protein and which crossreacts with the secondary antibodies used for staining. Staining in the embryo is seen in distinct punctate, about 1 μm, clusters at the axonal/glial interface in axon bundles (Fig. 1C-E), muscle precursors (not shown) and a pair of neurons (not shown). Two large bundles of axons connect each ventral nerve cord ganglion and contain a central core of axons surrounded by glial cell. The glial nuclei are on the outer surface and their processes wrap around the outside of the axonal bundle (Lane, 1981). Staining for HTL is seen mostly at the neuronal glial interface (Fig. 1D,E). This striped HTL staining pattern at the glial/axonal interface is generally unilateral and discontinuous.

Axonal glia in the 55% grasshopper embryo can be easily visualized on the connective axonal bundles between ganglia of the ventral nerve cord. Intact ventral nerve cords can be dissected from the embryo and cultured for over 24 hours (Condron and Zinn, 1997). Axonal glia retain their distinct elongated and flattened shapes in these cultured ventral nerve cords as they enwrap axons (Fig. 2A-F). The fly equivalent of these glia express HTL protein and mRNA (Gisselbrecht et al., 1996; Shishido et al., 1997). Staining grasshopper connective glia for HTL shows that these cells also contain the protein (Fig. 1D). In the fly, HTL activation results in the phosphorylation of MAPK (Gabay et al., 1997; Gisselbrecht et al., 1996). Staining the connective glia for activated MAPK reveals a stripe of staining (Fig. 2B,D) at the axonal glial interface that is very similar to the pattern of HTL protein (compare Fig. 1E with 2D). Unfortunately, it has not been possible to achieve histochemical conditions that allow double staining for these two antibodies. Staining for activated MAPK becomes more generalized throughout glial cells upon addition of FGF2 to the bath (data not shown). In addition, the normal flattened glial cells become more rounded after bath application of FGF2 (see below).

Glial nuclei in the grasshopper can be specifically stained with the antibody directed against the fly rk2/repo gene product (Halter et al., 1995). This rk2 nuclear staining reveals the normal flat morphology of the connective glia as their processes enwrap axons (Fig. 2E,F). These glia become rounded after 30 minutes of treatment with bath applied FGF2 (Fig. 2G,H). Glial morphologies in grasshopper ganglia treated with bath application of FGF2 are extremely similar to those of fly mutants that lack the FGF receptor (Gisselbrecht et al., 1996; Shishido et al., 1997). One explanation for this is that axons deliver a polar FGF signal necessary for the flattened morphology. Lack of the FGF receptor or uniform delivery of the FGF signal would interfere with the normal spatial function of this signalling pathway.

Activation of the HTL FGF receptor on axonal glia by bath application of FGF2 indicate the importance of delivering the signal in a spatially restricted manner. This can be achieved experimentally by immobilizing FGF2 on the surface of a polystyrene bead. To see what affinity glia have for an immobilized signal, FGF2-coated beads were placed into parts of the ganglia normally devoid of glia. One such area, the posterior-lateral portion of a ventral nerve cord ganglion, which contains mostly neuronal cell bodies, is shown in Fig. 3A. Insertion of an FGF2-coated bead into this region attracts a large number of glia which are normally absent from this part of the CNS (Fig. 3B). To test whether mature axonal glia, such as those shown in Figs 1 and 2, also respond to an FGF2-coated surface, beads were stuck to the ends of electrodes and held against these cells. Fig. 3C-F shows a sequence from a time lapse recording in which the glial cell immediately adjacent to the bead begins to round up and attach to the bead. The bead fell off the electrode after 60 minutes but still remained attached to a connective glial cell at 120 minutes. This indicates that mature axonal glia will respond to surface immobilized FGF2. Of 25 such electrode-presented beads, 19 induced rounding of the adjacent glial cell. It is not clear from the results of the experiment shown in Fig. 3B whether any cell division was involved in generating the glia that were attached to the bead. However, it was clear from the experiment depicted in Fig. 3C-F that no cell divisions were involved.

To examine the specificity of the bead coating for glial adhesion, 10 beads for each substrate were ‘floated’ (see Materials and Methods) onto a ganglion and the migration of glia was documented by time lapse microscopy (Fig. 4C-E). In addition, FGF2 was added to the bath and its potential counteractive effects on glial adhesion monitored (Fig. 4B,E). Ganglia were fixed and stained for the glial specific marker, rk2, to establish the identity of the attached cells (Fig. 4F). Data for bead coatings are summarized in (Fig. 4A). The data presented are the number of beads that acquired attached cells within the 30 minutes recording period. All of the cells that attached to the beads in this experiment were later identified as glia by rk2 staining. Beads coated with BSA, heat-denatured FGF2, IgG, grasshopper yolk protein or the antiserum directed against the intracellular portion of the fly HTL protein do not attract any glial cells. Uncoated beads, or those coated with FGF2, antiserum directed against the extracellular portion of the fly HTL protein or heparin do attract glia. The antiserum probably attracts glia by direct binding to the HTL receptor. This seems specific as antiserum directed against the intracellular portion of the same protein does not attract glial cells to the bead surface. Glia probably attach to uncoated beads by non-specific binding. Heparin, on the other hand, could induce glial attachment either non-specifically by perhaps not entirely coating the bead and thus leaving exposed non-specific binding regions, or perhaps more specifically. Heparin does bind and perhaps activates vertebrate FGF receptors (Green et al., 1996). In addition, heparin is present in the FGF2 mixture used in these experiments. As a further check of binding specificity, beads were followed by time lapse microscopy as they attached to glia and then their behavior was monitored upon addition of FGF2 to the bath. If the binding to the bead is FGF2 specific, then bath application of FGF2 should compete with the bead surface. An example of one such time lapse sequence is shown in (Fig. 4C-E). The bead in this case bound to the anterior lateral portion of a ganglion near the region where the connective axonal bundle forms. After 20 minutes, two glial cells have adhered to the bead (Fig. 4D, arrows). At this point, FGF2 was bath applied, to a final concentration of 38 ng/ml to see if it could compete with the bead-associated FGF2. 20 minutes after bath application of FGF2 (Fig. 4E), the glial cells, seen in Fig. 4D, have largely moved off the bead. Thus bath applied FGF2 can compete with bead-associated FGF2, indicating that it is indeed the FGF2 on the beads that induces glial adhesion. The effects of varying concentrations of bath applied FGF2 are plotted in (Fig. 4B). For each data point, 10 beads were floated onto a ganglion and glial migration onto the beads recorded by time lapse microscopy. Those beads with attached glia were scored after 30 minutes at which time FGF2 was added to the bath. The percentage glia remaining on beads after 60 minutes is indicated. Most glia moved off the beads after 38 ng/ml FGF2 was added to the bath. However, glia attached to beads coated with heparin or anti-HTL or uncoated were not detached by bath applied FGF2. This indicates that glia attach to beads coated with heparin or anti-HTL and uncoated beads in an FGF2-independent manner. All of the cells that remained attached to beads tested positive for rk2 staining (Fig. 4F). However, in separate experiments, glia attached to beads coated with FGF2 were stained for activated MAPK (Fig. 4G-H) as the fly HTL protein is thought to activate MAPK (Gabay et al., 1997). Interestingly only the closest cell shows strong MAPK activation (Fig. 4H).

In fruit flies mutant for the heartless FGF receptor, glia find axons but fail to enwrap them. As a result, the axonal connective bundles are decorated with round glial cells instead of cells with the normal flattened morphology as they enwrap axons (Gisselbrecht et al., 1996; Shishido et al., 1997). This indicates that FGF signalling is an important part of axonal glial morphogenesis in the fly. As a complement to these fruit fly genetic experiments, a set of cellular experiments were performed in the grasshopper. In the experiments described above, a very similar phenotype can be induced in mature glia in the grasshopper by bath application of FGF2. From both western blot analysis and the embryo expression pattern, it is very likely that FGF2 functions by activating the grasshopper homolog of the fly heartless FGF receptor. These experiments indicate that FGF signalling induces glial morphogenesis such that glia enwrap axons. Loss of the FGF receptor in the glia or delivery of an FGF ligand from all sides results in the functional loss of the morphogenetic signal. Thus it is the polarity of the FGF signal that is critical for its in vivo function in glial morphogenesis. It is unlikely that the glial adhesion seen in the majority of these experiments is due to an increase in proliferation. In the majority of these experiments, glia move onto beads in less than 5 minutes and no mitotic asters have been seen in the neighborhood of the beads. The overnight incubated bead experiment described in (Fig. 3A,B) however may involve glial proliferation due the longer time used for the experiment. While grasshopper axonal glial morphology can be experimentally manipulated by ectopic FGF, the means by which FGF signalling controls cell morphology is less clear. Cells can adhere to FGF2-coated beads (Fig. 3). However, it is not at all clear if glia adhere to beads by direct adhesion through FGF2 and the htl FGF receptor or whether glial shapes are perhaps ‘frozen’ by the signal. In this later model, glia would change shape in a random manner but locally stop changing in response to a polar FGF signal. Bath-applied FGF2 can out-compete bead associated FGF2 (Fig. 4B-E) indicating that if glia are adhering to beads directly via the receptors, then the bead-FGF2 receptor linkage is weak in that bath-FGF2 still has access to the bound receptors. A more likely scenario then is that polar FGF signalling affects the glial cytoskeleton and fixes the cellular morphology. Alternatively, FGF signaling might affect other adhesion molecules, as has been documented (Richard et al., 1995; Williams et al., 1994), however it is unclear how other adhesion molecules would bind to FGF2-coated beads. Binding to non-specific sites might just be enhanced upon FGF receptor activation. It should be noted that the amount of bath applied FGF2 required to half maximally displace bead-associated glia (Fig. 4B) is about 5 ng/ml. This would be roughly interpreted to mean that the amount of FGF2 on a bead surface is equivalent to 5 ng/ml bath-applied FGF2. While 38 ng/ml bath-applied FGF2 induces clear glial rounding on axons (Fig. 2G,H), it was difficult to clearly quantitate the effects of lower doses of FGF2, in contrast to glial migration off beads (Fig. 4B). Thus the concentration of bath FGF2 required to induce cell rounding is possibly similar to that on the bead.

The receptor localizes to a very specific part of the axonal glial interface. Drawing parallels from htl expression in the fly, where mRNA is found in axonal glia (Gisselbrecht et al., 1996; Shishido et al., 1997), this staining likely represents protein in the glia. However, the response of glia to bath-applied FGF2 in both morphological changes and a general increase in activated MAPK, argues that some lower undetectable amounts of protein probably exist distributed throughout glial surfaces. Although the glial cell depicted in Fig. 3C-F is likely to have HTL protein concentrated to the axonal glial interface as shown in Fig. 1E, it can still respond to an FGF2-coated bead. Therefore, the specific localization of the FGF receptor seen in Fig. 1E is probably a representation of the largest quantities of protein and might be an additional aspect of the polar pattern of this signalling pathway.

There is significant evidence that FGF signalling influences many other morphogenetic events. In invertebrates, FGF signalling is associated with sex myoblast migration in C. elegans (Burdine et al., 1998) and in Drosophila, with midline glial migration (Klambt et al., 1992), tracheal morphogenesis (Sutherland et al., 1996) and mesoderm and glial morphogenesis (Gisselbrecht et al., 1996; Shishido et al., 1997; Carmena et al., 1998; Michelson et al., 1998). Likewise, there is extensive evidence that FGFs are involved in vertebrate cellular morphogenesis, particularly with limb and lung development (Martin, 1998). In particular, developing mouse lung endodermal cells are attracted and attach to FGF-coated beads in a manner very reminiscent of the experiments described above (Park et al., 1998). There is also some evidence that FGF signalling influences vertebrate oligodendrocyte migration and morphology (Harari et al., 1997; Osterhout et al., 1997), as these glial cells mature to enwrap axons.

These experiments indicate that there is an FGF-like ligand in the insect made by or at axons which induces glial morphogenesis. The one known FGF-like molecule in the insect is branchless (Sutherland et al., 1996), which is a ligand for the breathless-encoded FGF receptor and is unlikely to be the ligand for HTL in this case. However, it is possible that FGF receptor activation in the grasshopper glia occurs in an FGF ligand-independent manner (Doherty and Walsh, 1996). The critical aspect of this FGF-signalling pathway is that the asymmetric presentation of the signal induces the morphologically correct cell shape change while general presentation does not (Fig. 5). This indicates that there is a complex subcellular component to the spatial aspects of initial axon-glial signalling.

I would like to thank Ben Barres, Jim Mandel, Alan Michelson, Os Steward, Scott Vandenberg, Kai Zinn and T. H. Pocket for excellent discussions concerning this work and Sandra Won and Lin Mei for critical advise concerning immunoprecipitation. I would also like to thank Os Steward, Ray Keller and Jay Hirsh for reading the manuscript. The idea to use the bead at the end of an electrode was Os Steward’s. Special thanks go to M. L. L. and B. P. C. This work was supported by NIH grant NS37223.