Peripheral Glia Direct Axon Guidance across the CNS/PNS Transition Zone

Transcription

1 Developmental Biology 238, (2001) doi: /dbio , available online at on Peripheral Glia Direct Axon Guidance across the CNS/PNS Transition Zone Katharine J. Sepp, Joost Schulte, and Vanessa J. Auld 1 Department of Zoology, University of British Columbia, Vancouver, BC, V6T 1Z4, Canada CNS glia have integral roles in directing axon migration of both vertebrates and insects. In contrast, very little is known about the roles of PNS glia in axonal pathfinding. In vertebrates and Drosophila, anatomical evidence shows that peripheral glia prefigure the transition zones through which axons migrate into and out of the CNS. Therefore, peripheral glia could guide axons at the transition zone. We used the Drosophila model system to test this hypothesis by ablating peripheral glia early in embryonic neurodevelopment via targeted overexpression of cell death genes grim and ced-3. The effects of peripheral glial loss on sensory and motor neuron development were analyzed. Motor axons initially exit the CNS in abnormal patterns in the absence of peripheral glia. However, they must use other cues within the periphery to find their correct target muscles since early pathfinding errors are largely overcome. When peripheral glia are lost, sensory axons show disrupted migration as they travel centrally. This is not a result of motor neuron defects, as determined by motor/sensory double-labeling experiments. We conclude that peripheral glia prefigure the CNS/PNS transition zone and guide axons as they traverse this region Academic Press Key Words: glia; axon; migration; pathfinding; Drosophila; grim; motor; sensory; ablation; gcm; peripheral. INTRODUCTION Glial cells interact with neurons in a variety of ways over the course of development. In the central nervous system (CNS), glia are known to function as permissive substrates over which growing axons migrate (Silver et al., 1982, 1987; Hidalgo and Booth, 2000). Some CNS glia express a variety of molecules such as Slit and Netrin, which actively guide migrating neurites to their final targets (reviewed by Jacobs, 2000). Because axonal guidance molecules, which are expressed by glia, are conserved through evolution from invertebrates to vertebrates, the role of glia in generating the pattern of a large and complex nervous system is highly significant. During later nervous system development, when glial ensheathment of axonal bundles occurs, communication between glia and neurons is also apparent. Absolute numbers of glia are influenced by the amount of nerve bundles available for wrapping. The numbers of glia can be adjusted through apoptosis in mutants that have ectopic or missing axons (Sonnenfeld and Jacobs, 1995a). 1 To whom correspondence should be addressed at University of British Columbia, Department of Zoology, 6270 University Blvd., Vancouver, BC, V6T 1Z4, Canada. Fax: (604) auld@zoology.ubc.ca. Conversely, CNS neurons undergo apoptosis when glia are genetically removed (Booth et al., 2000). This suggests that glial cells participate in trophic interactions with neurons over the course of development, which results in the generation of an efficient, functioning adult nervous system. The roles of glial cells in CNS axonal pathfinding are well established and are analogous between vertebrates and invertebrates (reviewed by Auld, 1999). In contrast, the function of glia in generation of the peripheral nervous system (PNS) pattern is not well understood in any organism. This has been largely the result of the complexity of vertebrate morphology and genetics and, in the simpler invertebrates, a lack of cellular markers. We have recently generated strains of Drosophila melanogaster that can be used to generate PNS glial-specific gene expression using the GAL4/UAS system (Brand and Perrimon, 1993; Sepp and Auld, 1999). These lines are useful as strong cellular markers and have lent themselves to a full characterization of peripheral glial growth over the course of embryonic and larval development (Sepp et al., 2000). The development of peripheral glia in the Drosophila embryo is very similar to vertebrate PNS glial development. Peripheral glia first appear at the lateral edge of the CNS and migrate into the periphery along growing axons (Fig. 1) /01 $35.00 Copyright 2001 by Academic Press All rights of reproduction in any form reserved. 47

2 48 Sepp, Schulte, and Auld Initially, peripheral glia are compact and oval-shaped. As they mature, they extend cytoplasmic processes, which form a single wrap around PNS axon fascicles, similar to nonmyelinating Schwann cells (Jacobs and Goodman, 1989; Auld et al., 1995; Sepp et al., 2000). The differentiation of peripheral glia is controlled by transcription factors that have relatives in vertebrates such as Krüppel/Krox-20, Gcm/GCMa/b, and Pointed/c-Ets-1/2 (Hedergen et al., 1993; Klaes et al., 1994; Hosoya et al., 1995; Jones et al., 1995; Albagli et al., 1996; Romani et al., 1996; Vincent et al., 1996; Kim et al., 1998; Reifegerste et al., 1999). Furthermore, fully differentiated peripheral glia of Drosophila and vertebrate glial cells express related cell adhesion molecules such as Neuroglian/Liter1, Gliotactin/Neuroligin, and Connectin/LIG-1 (Moos et al., 1988; Bieber et al., 1989; Auld et al., 1995; Suzuki et al., 1996; Raghavan and White, 1997; Gilbert et al., 2001). Anatomical characterization of development in both vertebrates and Drosophila suggests that peripheral glia function as intermediate targets for migrating neurites as they cross the CNS/PNS transition zone (TZ). During early Drosophila neurodevelopment, peripheral glia are situated at the lateral edge of the CNS and are arranged in a compact funnel-shaped array. Motor and sensory neurons make growth cone contacts with the glia as they exit and enter the CNS, respectively. The pathways along which early pioneer axons travel in this region appear to accommodate the physical arrangement of the peripheral glia (Sepp et al., 2000). As the peripheral glia migrate into the PNS, they extend cytoplasmic processes and form tubes through which sensory axons migrate into the CNS. Thus peripheral glia are a substrate for sensory axon migration in the distal to medial region of the periphery. Also, the glia obscure other possible sensory axon migrational substrates such as muscle, with their cell processes. In vertebrate embryonic development, Boundary Cap cells demarcate the position on the spinal cord where future motor and sensory neurons migrate across the TZ. Boundary Cap cells express Schwann cell markers and are derived from the neural crest (Nieder- FIG. 1. Schematic diagram of early embryonic neurodevelopment. Motor neurons (red), sensory neurons (orange), and peripheral glia (blue) are indicated. Anterior is to the top, posterior to the bottom. CNS is on the left. Peripheral glia arise from the CNS and have also been called exit glia, which refers to an early, transient stage in their development (Klämbt and Goodman, 1991). (A) Stage 12/1. Pioneer motor neuron acc growth cones approach peripheral glia that arise at CNS/PNS border. The first sensory cell appears in the PNS. (B) Early stage 13. Peripheral glia proliferate and acc growth cone exits the CNS. Further sensory cells appear. (C) Mid stage 13. acc growth cone reaches lateral chordotonal region in lower segment. Further motor and sensory neurons arise. Sensory axons from dorsal region begin to connect with motor axons (lower segment). Peripheral glia migrate into the PNS along axonal fascicles. (D) Stage 14. acc growth cone has reached sensory axon projections in all segments to form a continuous motor/sensory fascicle. Further sensory neurons arise in the PNS and project axons centrally through a cone-shaped array of peripheral glia located at the TZ. Peripheral glia continue to migrate into the PNS. (Adapted from Casey Kopczynski, unpublished data.)

3 Developmental Function of PNS Glia 49 FIG. 2. Pattern of peripheral glia during early embryonic stages. Glia are stained with anti- -galactosidase (blue); neurons (brown) are labeled with MAbs 22C10 (A, B, D) and 1D4 (C). Anterior to the top, CNS is to the left (B, D); CNS is in the middle (A, C). Magnification is higher in B and D. (A) Cytoplasmic staining of rq286::tau-lacz embryo shows labeling of compact peripheral glia at stage 12/1. acc pioneer neuron makes growth cone contact with peripheral glia (arrows). (B) rq286 nuclear enhancer trap staining at stage 14 shows glial cell bodies migrating into the periphery. Peripheral glia migrate along the anterior fascicle (lower arrow) and lower fascicle (upper arrow). (C) rq286 nuclear enhancer trap staining at stage 13 shows motor neuron pioneers migrating through cone-shaped arrays of peripheral glia (arrows). (D) Corresponding cytoplasmic labeling of rq286::tau-lacz embryos at same stage as B shows glial processes extending along anterior and posterior fascicles (upper and lower arrows, respectively), which are kept spatially distinct from one another. For a more detailed description of peripheral glial development, see Sepp et al. (2000). länder and Lumsden, 1996; Golding and Cohen, 1997). From analysis of growth of dorsal root ganglia (DRG) neurons cultured on dorsal root/spinal cord cryosections, it is suggested that Boundary Cap cells attract the central migrations of DRG axons into the spinal cord (Golding and Cohen, 1997). We have used Drosophila to further explore the role of peripheral glia as mediators of axon guidance by carrying out peripheral glial ablations during early neurodevelopment and by analyzing the glial differentiation mutant, glial cells missing (gcm) (Hosoya et al., 1995; Jones et al., 1995; Vincent et al., 1996). In the absence of peripheral glia, sensory axons migrate into the CNS at erratic positions along the neuraxis. In their migrations toward the CNS, sensory axons stall and exhibit pathfinding defects. Therefore, peripheral glial processes may act as a physical barrier, which prevents sensory axons from recognizing inappropriate migrational pathways. Loss of peripheral glia also leads to abnormal migration of pioneer motor axons across the TZ. Thus the initial trajectories of pioneer motor axons may be influenced by peripheral glia. During later developmental stages however, motor axons are capable of migrating to their correct muscle targets without peripheral glia. The results support previous anatomical and in vitro studies of the TZ and show that peripheral glial cells play an important role in neuronal pathfinding. MATERIALS AND METHODS Fly Strains and Genetics The rq286 nuclear lacz enhancer trap line (Fig. 2A) and the GAL4 drivers rq286:gal4 (rq286#11) and rq14:gal4 (rq14#7) were generated and characterized previously (Klämbt and Goodman, 1991; Sepp and Auld, 1999; Sepp et al., 2000). The repo:gal4 driver was generated by a random GAL4 P element insertion into the repo locus and its expression matches the endogenous Repo expression pattern (S. Benzer, unpublished data). Apoptotic UASced-3#6-6 and UAS-grim transgenics were used for previous ablations (Shigenaga et al., 1997; Zhou et al., 1997). Cell marker strains UAS-tau-lacZ (Hidalgo et al., 1995) and UAS-lacZ were provided by Andrea Brand and Bloomington Stock Center, respectively. The mutant glial cells missing (gcm P1 ) is a null deletion mutant generated by imprecise P-element excision (Jones et al., 1995). All GAL4 drivers were crossed to the UAS-grim/ced-3 lines to produce doubly heterozygous experimental embryos. All drivers and apoptotic lines (as heterozygotes) have wild-type nervous system patterning compared to that of wild-type Oregon R embryos.

4 50 Sepp, Schulte, and Auld Immunohistochemistry Embryos raised at 25 C were stained using immunohistochemical techniques as reported previously (Halter et al., 1995), except for mab 49C4 staining. For labeling with mab 49C4 (IgM), collected embryos were dechorionated in 50% household bleach and rinsed with distilled water. A fixative solution of 6% HgCl 2 /0.12% NaOAc diluted 1:1 with 10% formaldehyde was prepared. A 5-ml aliquot of the fixative was placed in scintillation vial with 5 ml n-heptane and embryos were fixed for 20 min with nutation. Standard embryo staining techniques (Halter et al., 1995) were used following the fixation. Mercury fixation disrupted many tissues as well as the Repo epitope, thus double labeling of neurons and glia could not be done using mab 49C4. Neurons were labeled using mouse mabs 1D4 and 22C10 at 1:2 dilution (Davis et al., 1997; Fujita et al., 1986), rabbit anti-hrp at 1:300 (Jackson ImmunoResearch), and mouse mab 49C4 at 1:5 dilution. The mab 49C4 labels an unknown antigen on lateral chordotonal sensory neurons (Kolodziej et al., 1995). Glial nuclei of rq286 (wild-type) embryos were detected with rabbit anti- galactosidase at 1:1000 (Cappel). Glial nuclei were otherwise detected using rabbit anti-repo at 1:200 (Halter et al., 1995). For glial cytoplasmic labeling, a goat anti-rabbit-ap secondary (Jackson ImmunoResearch) was used at 1:1500. For detection of neurons, a horse anti-mouse-biotin secondary was used with AB reagent for HRP reactions (Vectastain ABC Kit, Vector Laboratories) according to the manufacturer s instructions. Embryos were cleared in a series of 50, 70, and 90% glycerol in PBS. Dissected embryos were mounted using 90% glycerol in PBS. Embryo PNS segments were photographed on Fujichrome RTPII and Kodachrome ISO 64 film using a Zeiss Axioskop. Images were digitized and processed using Adobe Photoshop 5.5. For confocal microscopy, embryos were incubated in 1:200 dilutions of fluorescent secondaries Alexa Fluor 488 and Alexa Fluor 568 (Molecular Probes). Fluorescently labeled embryos were mounted with Vectashield (Vector Laboratories) and analyzed using a Bio-Rad Radiance Plus confocal microscope. Images were analyzed using NIH Image 1.62 and Adobe Photoshop 5.5. RESULTS Ablation of Peripheral Glia From analyzing peripheral glial development in wild-type embryos and larvae (Sepp et al., 2000), we hypothesized that peripheral glia could have roles in axonal guidance. Motor axon growth cones have been previously shown to contact the nerve root glia (Segment Boundary Cell) in the CNS before turning toward the periphery (Bastiani and Goodman, 1986; Jacobs and Goodman, 1989). The growth cones then migrate onto the peripheral glia, which are born at the lateral edge of the CNS (Figs. 1A, 1B, 2A, and 2C). After contacting the peripheral glia, the pioneer motor axons finally migrate into the periphery. As well, sensory axons originating in the periphery migrate toward the CNS through a tube of peripheral glial cellular processes, which forms in the ventral to lateral regions (Figs. 1C, 1D, 2B, and 2D). Therefore, during a significant portion of sensory axon migration, the neurites are obscured from migrational substrates other than peripheral glia or motor axons, which are also encircled by the glia. Given the close physical contacts peripheral axons make with the peripheral glia, it is likely that glia could play a role in guiding axons to help form the mature nervous system pattern. To address this hypothesis, we ablated peripheral glia early in embryonic development and assessed the resultant effects on motor and sensory axon pathfinding. As well, we assessed the neuronal phenotype of the glial differentiation mutant, glial cells missing (gcm) (Hosoya et al., 1995; Jones et al., 1995; Vincent et al., 1996). For ablations, we used the GAL4/UAS targeted gene expression system of Drosophila, where peripheral glialspecific GAL4 lines were used to drive the expression of the apoptotic genes grim and ced-3 (Brand and Perrimon, 1993; Shigenaga et al., 1997; Zhou et al., 1997). Grim is a novel gene (Zhou et al., 1997) that stimulates apoptosis in Drosophila and CED-3 is a C. elegans caspase (Shigenaga et al., 1997). The glial line rq286:gal4 (Fig. 2; see also Sepp et al., 2000), which was used to drive expression of the cell death genes, has strong expression in all peripheral glia at the beginning of gliogenesis. The rq286:gal4 enhancer trap is expressed in only the peripheral glia and not the Repo-positive glia associated with sensory organs, nor the ISN and SN glia, which ensheathe the peripheral nerves at the most proximal region of the peripheral nerves within the CNS (Ito et al., 1995; Sepp et al., 2000). Three other fly lines were used to verify that ablation effects were attributed to the loss of peripheral glia and not to other tissues. The rq14:gal4 line was used as a negative control. Even though it is an independent GAL4 insertion in the same gene as rq286, it has very weak GAL4 expression in a small subset of peripheral glia. We also used the glial-specific repo:gal4 driver for ablations. The repo gene is expressed in all ectodermally derived glia within the embryonic PNS and CNS (Xiong et al., 1994; Campbell et al., 1994; Halter et al., 1995). In addition we analyzed the PNS phenotype of the gcm null mutant, which lacks differentiated glia (Hosoya et al., 1995; Jones et al., 1995; Vincent et al., 1996). The PNS phenotype of the gcm mutant was not previously analyzed in detail with regard to sensory axons. The gcm mutant also served as a control, to verify that migrational defects were ascribed to glial deficiency and not to cell debris from the ablated glia. By generating embryos carrying the rq286:gal4 insert to drive UAS-grim or UAS-ced-3 (referred to herein as rq286::grim/ced-3) we observed a spectrum of nervous system phenotypes in embryonic hemisegments, which ranged from having disrupted CNS and PNS pathways to being essentially normal. The neuronal defects mirrored the extent of glial ablation, which varied from complete removal of glia to wild type. The presence or absence of glia was assayed three different ways. First, we included a UAS-lacZ marker in the background of our mutants. Loss of anti- -galactosidase staining in the peripheral glia was one indicator of glial loss. We also used an antibody to the Repo protein (Xiong et al., 1994; Campbell et al., 1994; Halter et al., 1995), to detect any peripheral glia remaining in the embryo. However, weakened Repo staining could

5 Developmental Function of PNS Glia 51 also be associated with glial cells that were possibly still present and only in the process of dying. Such embryos could potentially still provide axonal guidance cues on their membranes. Therefore, our final and possibly the most reliable verification of glial loss was microscopic analysis of the embryo using Nomarski optics, to verify that no glial sheaths were present around the peripheral nerves. Such embryos that lacked a visible glial sheath showed defasciculated peripheral nerves as well as axonal pathfinding defects, as described below. Because GAL4-induced gene expression in a given tissue can be heterogeneous (Lee and Luo, 1999), a high enough level of apoptotic gene expression to cause glial death did not always occur in all segments. Also, temporal differences in activation of targeted cell death were likely to have occurred within peripheral glia of a given segment as well as across all segments of the embryo. If peripheral glia are ablated earlier, the resultant neuronal phenotype would likely be far more severe compared to that of a later glial-ablated embryo; however, both could show equal glial loss at late stages. For quantitative analysis, gcm mutants that showed even disruption of glia across all segments provided the most consistent phenotypes in terms of severity. Both moderately and more severely glial-ablated embryos, although variable in severity, were qualitatively similar to each other and to the gcm mutant. Thus the moderate and severely glial-ablated embryos were scored together. The ablated glia phenotypes generated by grim expression were identical to those generated by ced-3 expression. Only glial-ablated embryos generated with UAS-grim are shown here. rq286::grim embryos with highly aberrant nervous system patterning correlated with early evidence of glial ablation. From stage 13 onward, when the peripheral glia normally would have migrated into the periphery (Figs. 1 and 2), the nuclei of the peripheral glial cells in these embryos remained inside the CNS (Figs. 3B and 3D, arrowheads). These glial nuclei were observed at the lateral edge of the CNS and had patchy and faint Repo staining, suggesting that they may be pyknotic. The peripheral nerves associated with the ablated glia were always defasciculated (Figs. 4B, 4D, 5B, Table 1) in contrast to wild-type nerves, which never have defasciculation. Outside the lateral border of the CNS, macrophages could be seen with Nomarski optics and sometimes small, blue-stained (Repo-positive) fragments of nuclei could be observed within the macrophages, suggesting that the apoptotic peripheral glia were later removed from the CNS and processed by macrophages (data not shown). Similarly, repo::grim embryos showed loss of peripheral glia in the PNS associated with disrupted PNS nerve pathways (Figs. 3D, 3G, 3H, and 4D). In embryos at later stages, peripheral glial cells in the PNS were not visible with Repo staining nor were outlines of these cells observed using Nomarski optics. Therefore, other CNS glia not targeted for ablation were unable to migrate to the PNS and compensate for peripheral glial loss. However, it is likely that peripheral glial ablation could have non-cellautonomous effects on other CNS glia, given that glial cells actively participate in the removal of other apoptotic cells out of the CNS (Sonnenfeld and Jacobs, 1995b). The physical removal of apoptotic glia would likely be carried out by neighboring glial cells, which would in turn displace them from their usual positions. Given that most other CNS glia play roles in prefiguring CNS axon pathways, any glial misplacement could cause disruptions of axon migration within the CNS (Jacobs and Goodman, 1989; Auld, 1999). Therefore, we focused our analysis on axon migration within the PNS, where only the peripheral glia ensheathe nerve bundles. Sensory Neuronal Development in the Absence of Peripheral Glia The effects of peripheral glial ablation on sensory neuronal development was assessed by staining embryos with the mab 22C10. This monoclonal recognizes Futsch, a novel microtubule-associated protein that is necessary for axonal and dendritic growth (Hummel et al., 2000; Roos et al., 2000). Glial-ablated rq286::grim and repo::grim embryos as well as gcm embryos showed generally weaker labeling of sensory axons with anti-futsch/22c10. Such loss of neuronal mab 22C10 antigenicity was also previously noted in the glial differentiation mutant pointed (Klaes et al., 1994). Sensory neurons project axons along two main peripheral nerves into the CNS: the anterior fascicle (af) and the posterior fascicle (pf) (Fig. 3A). The sensory neurons in the negative control rq14::grim embryos stained in a wild-type pattern. In the rq286::grim glial-ablated embryos, sensory axons displayed a range of errors including stalls, improper fascicle choice, defasciculation, and misguidance on muscle. These effects were qualitatively similar to those of the gcm and repo::grim mutant phenotypes. Sensory Neurons Stall in the Absence of Peripheral Glia In both rq286::grim and repo::grim glial-ablated and gcm abdominal hemisegments, we observed sensory axons stall in their approach to the CNS (Figs. 3B 3E, arrows; 3F, arrowhead). In stage 14 gcm embryos, stalling could be accurately scored, since the few mab 22C10-positive axons that emanated from the CNS were weakly stained, allowing the leading edge of the anterior and posterior fascicles to be visible (compare left sides of Figs. 3A and 3C). We also observed evidence of stalling, in which one segment in an embryo was more delayed compared to another segment beyond the normal range of segment-to-segment variation. Stalls in stage 14 gcm embryo abdominal segments A1 A6 were observed in 30% of hemisegments (n 94). In more mature stage 16 embryos, stalling could not be as accurately scored, since young migrating sensory axons at later stages fasciculate with the older sensory axons that had already migrated much further into the CNS. Therefore, the only definite stalls were those in which most axons of a fascicle had stalled together in the same region of the embryo (Figs.

6 52 Sepp, Schulte, and Auld FIG. 3. Sensory neurons stall in the absence of peripheral glia. Sensory neurons are labeled with mab 22C10. Glial nuclei are labeled with anti-repo in A, B, D, and F H. CNS/PNS transition zone is indicated with a vertical bar. There are almost no Repo-positive glia in gcm mutants and the few Repo-stained cells that do appear do not strongly correlate with less aberrant axon tracts. In glial-ablated embryos, Repo staining reveals some residual glial nuclei (i.e., arrowheads B, D). Anterior is to the top, CNS is to the left. (A) Wild-type late stage 16 embryo shows anterior fascicle (af) and posterior fascicles (pf) bundled tightly together, yet distinct, at the TZ (asterisk). (B) An rq286::grim embryo shows sensory neurons of anterior fascicle stalled at the TZ (arrow). The anterior fascicle in the same focal plane as the neighboring glia is also bifurcated. Arrowheads indicate weakly stained nuclei (compare to A). (C) In a stage 15 gcm embryo, the anterior fascicle of the top segment, compared to the bottom segment, is well advanced toward the CNS, indicating delayed migration (compare arrows). (D) In a repo::grim-ablated embryo, all ectodermally derived glia, which include the peripheral glia as well as sensory organ-associated glia (see Halter et al., 1995), have been targeted for ablation. The anterior fascicle shows stalling in the lateral region of the embryo (arrow) and a sensory projection migrating to the neighboring anterior segment (concave arrow). (E) A gcm stage 16 embryo shows incorrect migration of anterior fascicle neurons to an adjacent hemisegment (concave arrow). The v and v sensory cluster axons did not fuse together to form the posterior fascicle in the lower segment. (F) A rq286::grim glial-ablated embryo shows overall misplacement

7 Developmental Function of PNS Glia 53 FIG. 4. Peripheral glial loss is associated with peripheral nerve bundling defects and defasciculation. Embryos are stained with mab 22C10; anterior is to the top; CNS is to the left. CNS/PNS transition zone is indicated with a vertical bar. Magnification is higher in B and D than in A and C. (A) The anterior and posterior fascicles (upper and lower arrows, respectively) in wild-type embryos at stage 16 are distinct, yet closely bundled together by peripheral glia at the TZ (asterisk). (B) An rq286::grim peripheral glial-ablated embryo shows wide separation of anterior and posterior fascicles (asterisk). The sensory axons appear defasciculated (concave arrows). Axons traverse incorrectly from the posterior to the anterior fascicle (solid arrow). (C) The gcm embryo has aberrant nerve fascicles at the TZ in all segments, which can range from loss of anterior and posterior fascicle distinction (top asterisk) to wide separation of these fascicles (bottom asterisk). Pathfinding defects manifest as incorrect transit of axons from the anterior fascicle to the posterior fascicle (top arrow) or vice versa (bottom arrow). (D) A repo::grim glial-ablated embryo shows incorrect migration of axons from the posterior fascicle to the anterior fascicle (arrow) and defasciculation and wide separation of anterior and posterior fascicles (asterisk) and nerve defasciculation. Concave arrows show defasciculation. 3B, 3D 3H). In stage 16 gcm embryos, such stalls were observed in 7% (n 116) of abdominal hemisegments. Compared to the wild type, control rq14::grim embryos did not show any stalling of sensory axons. In rq286::grim and repo::grim embryos, segments that showed evidence of glial loss also showed sensory neuronal stalls (Figs. 3B, 3D, 3F 3H) at 76% (n 112) and 85% (n 96), respectively. It is possible that the defective glial differentiation in the gcm mutant may not be severe enough to completely remove all glial-derived function. Indeed, there is also a second gcm motif gene, gcm2, in the Drosophila genome (Akiyama et al., 1996), which likely has overlapping function with gcm. This may account for the higher penetrance of the stalling phenotype in glial-ablated embryos compared to gcm embryos. Alternately, growth cones encountering dead cell debris might lead to further stalling of neurites in glialablated embryos compared to that in gcm embryos. Sensory Axons Are Misguided in the Absence of Peripheral Glia In glial-ablated and gcm embryos, sensory axon pathways were misrouted. Of the hundreds of gcm and glial-ablated segments that were analyzed, we did not observe any of sensory cell bodies and stalling of dorsal cluster axons (arrowhead). The acc pioneer motor axon (arrow) is stalled in the middle segment, in which peripheral glia failed to migrate to the periphery. The dorsal cluster neurons of the same segment crossed incorrectly to the anterior segment (concave arrow). (G) A repo::grim embryo with weaker ablation than that in (D) shows dorsal cluster axons misrouted and stalled in the lateral region (solid arrow) and misrouting of dorsal cluster axons across a segment boundary (concave arrow). Arrowheads indicate ventral cluster neuron cell bodies that are misplaced. The ventral cluster neuronal cells do not associate with Repo-positive support cells in wild-type embryos. (H) A repo::grim embryo shows stalling of dorsal cluster axons in top two segments (arrows). In the middle segment, the acc motor pioneer is weakly labeled and appears stalled with respect to the lower segment. (D, E, F) Stalling of the sensory anterior fascicle is seen in the lateral region of the repo::grim, gcm, and rq286::grim embryos (asterisks).

8 54 Sepp, Schulte, and Auld TABLE 1 Summary of Motor and Sensory Neuron Defects in Mutants Lacking Peripheral Glia Mutant Phenotype scored Affected hemisegments/ total scored Percentage hemisegments affected A. Deviations of sensory axon staining of mature embryos compared to wild type (mab 22C10) rq286::grim Any deviation from wild type a 708/ Axon crossings between af and pf 59/ Axon crossing segment boundary 51/ Stalls at stage 16 85/ gcm Any deviation from wild type 396/ Axon crossings between af and pf 60/ Axon crossing segment boundary 1/101 1 Stalls at stage 14 28/94 30 Stalls at stage 16 8/116 7 repo::grim Any deviation from wild type 464/ Axon crossings between af and pf 61/96 63 Axon crossing segment boundary 56/96 58 Stalls at stage 16 82/96 85 wild type Any PNS axon defect 0/62 (0) B. Deviations of motor axon staining compared to wild type (mab 1D4) Mutant Phenotype scored Number of hemisegments/ total scored Percentage hemisegments affected rq286::grim Any deviation from wild type b 864/ Axon stalls at stage 13/14 44/ gcm Any deviation from wild type 444/ Axon stalls at stage 13/14 33/86 38 repo:grim Any deviation from wild type 132/ Axon stalls at stage 13/14 41/93 44 wild type Defasiculation/bundling 0/78 (0) a Deviations from wild type Oregon R sensory pattern included defasciculation in all cases, but could also include abnormal separation of nerves, axon misrouts, and stalls. b Deviations from wild type Oregon R motor pattern in all cases included defasciculation and abnormal separation of peripheral nerves at the TZ, which were compared to wild type axon fasciculation and bundling. segments that were entirely wild-type in sensory neuronal patterning compared to that of control (Table 1A). In contrast, mature wild-type embryos had a 0% frequency of sensory axon patterning defects in the PNS. For our analysis of axon pathfinding, we present embryos stained without the Repo antibody, which tends to obscure the axon phenotypes, especially in the ventral region. In the mutants, sensory axon migration errors included incorrect crossing of sensory neurons to neighboring segments (Figs. 3D 3G, concave arrows; Table 1A). Also, sensory axons were observed to incorrectly switch pathways between the anterior and posterior fascicles, compared to wild types, where these two fascicles are always distinct from each other (Figs. 4B 4D, solid arrows; Table 1A). The pathfinding of sensory axons across the TZ in all peripheral glial-deficient embryos was highly erratic (Fig. 4, asterisks), and often the sensory axons entered the CNS at incorrect positions. This phenotype could be the result of either a simple failure of tight bundling by the glia or a lack of guidance to the correct CNS entry position. In the most severely ablated rq286::grim and repo::grim embryos, greater PNS defects were observed alongside TZ defects. In these embryos, defects ranged from misplacement of sensory neuron cell bodies, disrupted axogenesis, and variable loss of sensory neurons (Figs. 3B, 3D, 3F, and 3G). It is unlikely that misplacement of sensory cell bodies causes all axon pathfinding defects, given that gcm mutants do not show such cell body misplacement, yet display all types of sensory pathfinding defects, as observed

9 Developmental Function of PNS Glia 55 FIG. 5. Projection of lateral chordotonal axons into the CNS after peripheral glial loss. Stage 16 embryos were labeled with mab 49C4. The mab 49C4 requires staining with mercury fixation, which leads to a roughened tissue morphology. Anterior is at top. CNS is on the left (A, B, D F) and in the middle (C). B, D, and F are at higher magnification. The boundary of the CNS is marked with vertical lines. (A) Four lateral chordotonal (lch) organs (brown) project axons into the CNS along the anterior fascicle (af) in wild-type embryos. A single ventral cluster sensory cell also projects an axon into the CNS along the posterior fascicle (pf). (B) A gcm mutant embryo shows lch fascicle bifurcation (asterisk) just outside the CNS/PNS border in the top segment. (C) Lch axons project incorrectly into the CNS of an rq286::grim-ablated glia embryo. In two hemisegments, neurites reversed direction and subsequently entered the CNS more posteriorly (arrows). Another segment shows erroneous posterior penetration of sensory neurons into the CNS (arrowhead). (D) A gcm embryo shows an lch axon turned posteriorly at the ventralmost border of the abdominal muscle field. The outlines of motor neurons projecting from the CNS can be seen with Nomarski (arrow). The lch neurons have fasciculated with the motor neurons up to this point. CNS/PNS boundary is indicated by vertical line at the bottom. (E) A higher magnification of the rq286::grim embryo in C. Top segment shows axons turned ventrally and posteriorly (concave arrow). Lch axons are highly defasciculated in the lower segment (solid arrow). (F) A glial-ablated rq286::grim embryo shows axons near the CNS at an incorrect focal plane. Axons appear to extend along the ventral oblique muscle edges, which normally reach slightly below the CNS. Dotted lines highlight the more ventral longitudinal muscle cell boundaries. in glial-ablated embryos. Furthermore, it is unlikely that the defects observed in gcm embryos are the result of glial cells differentiating into neurons. The ventral and ventral clusters of sensory axons are not associated with any GCMand Repo-positive support cells/glia (Xiong et al., 1994; Campbell et al., 1994; Halter et al., 1995; Hosoya et al., 1995; Jones et al., 1995; Vincent et al., 1996). Nevertheless, these neurons exhibit all the phenotypic classes observed in the lch and dorsal cluster axons in the mutants studied. Although all segments that showed evidence of glial ablation always had sensory neuronal defects, and all PNS segments of the gcm mutant appeared aberrant, we investigated to what degree the defects were attributable to errors in fasciculation, as opposed to a simple mislocation of the TZ in abdominal segments A1 to A6. For the first category, misroutings of sensory neurons were scored either if axons transferred erroneously between the anterior and posterior fascicles (i.e., bottom segment of Fig. 4D) or if the

10 56 Sepp, Schulte, and Auld FIG. 6. Effects of peripheral glial loss on early and late motor axon pathfinding. Motor axons were labeled using mab 1D4. Anterior is to the top. CNS is in the middle (A, C, E) and to the left (B, D, F). CNS/PNS transition zone is indicated with a vertical bar (B, D, F H). (A) Wild-type pattern of motor axon staining shows peripheral nerve pioneers exiting CNS at stereotyped trajectories (arrows) during early development (stage 13 late). (B) A wild-type mature stage 17 embryo shows stereotyped motor projections in the PNS. The ISNb nerve is indicated (arrow). (C) An early stage 13 rq286::grim embryo shows aberrant trajectories of motor neurons migrating into the PNS. Arrows indicate differences in the amount of motor axon extension into the PNS. (D) In the rq286::grim-ablated glia embryo, the basic motor axon pattern of both the CNS and PNS is retained. The ISNb branch is indicated (arrow). (E) A gcm mutant at late stage 13 shows irregular motor neuron trajectories in TZ region. Arrows indicate abnormal separation of ISN and SN nerves. (F) The gcm mutant has more significant disruption of peripheral nerve tract patterning than the glial-ablated embryo at stage 17. In the most severely affected (top) hemisegment, CNS longitudinal fascicles are weak and PNS ISNb nerve branch is lacking (arrow) but the general pattern is retained. (G) A wild-type embryo at stage 17 shows motor axons and glial nuclei (arrowheads) along the peripheral nerves. (H) A repo::grim embryo in which the PNS is devoid of Repo-labeled glial nuclei, yet some glia are labeled in the CNS (arrowheads). The motor tracts of the PNS are largely normal, with correct branching patterns. ISNb branch is indicated (arrow).

11 57 Developmental Function of PNS Glia FIG. 7. Sensory and motor axon patterning in the gcm embryo. For confocal microscopy of stage 16 embryos, sensory neurons were labeled with mab 22C10 (red) and both motor and sensory neurons were labeled with anti-hrp (green). Anterior is to the top; CNS (ventral) is to the left; dorsal is to the right. (A) A wild-type embryo shows sensory axons emanating from the ventral (v) and ventral (v ) clusters fusing together to form the posterior fascicle (pf). The v cluster axons first migrate dorsally (arrowhead) before fusing with the v cluster axons. The peripherally migrating ISNb motor branch exits the ISN tract close to where the v and v axons fasciculate together. The ISN motor branch is in a common fascicle together with the anterior fascicle in the ventral region of the embryo. Similarly, the SNa motor branch fasciculates with the posterior fascicle in the ventral region. (B D) In gcm embryos, sensory axons of the v cluster migrate into the CNS independently of the SNa motor fascicle (concave arrows). The v cluster sensory axons do not initiate their migrations by turning first dorsally (arrowheads, compare arrowheads with A). (C) In the upper segment, the anterior fascicle initially migrated along the combined motor sensory ISN/af tract in the lateral region of the embryo, yet deviated from it near the CNS entry point (downward arrow). The ISN motor root is identified by its association with the ISNb root in the ventral region. (B, C) The TZ region is very poorly fasciculated in the absence of glia (asterisks, compare to A). (D) A sensory projection from the v cluster neurons is stalled close to the CNS entry point (arrow). two fascicles fused together as one at any point within the PNS (i.e., top segment of Fig. 4C). The first category thus includes all obvious evidence of gross axon misguidance. The second category included all segments in which the anterior and posterior fascicles were entirely distinct (see top segment of Fig. 4D, for example). Of 101 hemisegments scored, 60 showed anterior/posterior fascicle crossings or fusions, whereas 41 hemisegments maintained distinct anterior and posterior fascicles. The latter category could still include sensory axon misroutings, given that some fascicles appeared to enter the CNS at positions very different from the wild-type entry point (i.e., bottom segment Fig. 4C). It is possible that entry to the CNS at the wrong position could cause further misguidance of the sensory neurons. Therefore, the resultant frequency of 59% (n 101) gross sensory axon misguidance in the PNS for the gcm mutant could underrepresent the actual frequency. To investigate more closely what happens to sensory

12 58 Sepp, Schulte, and Auld neuron pathfinding, we analyzed the development of a small subset of sensory neurons, the lateral chordotonals (lch) using the mab 49C4 (Fig. 5). The mercury fixation necessary for staining with the 49C4 antibody disrupts the embryonic tissue, such that most embryos do not stain well. Thus not enough mab 49C4-stained embryos were obtained to accurately quantify phenotypes. In wild-type and rq14::grim control embryos, these neurons migrated into the CNS along the anterior fascicle (af; Fig. 5A). In rq286::grim glial-ablated embryos, the lch neurites commonly stalled at the TZ and showed pathfinding errors as they approached the CNS (Figs. 5C and 5E). Very often, the axons headed posteriorly and entered the CNS incorrectly toward the posterior fascicle route (Fig. 5C, arrows). As well, the four lch axon projections showed significant defasciculation in some segments (Fig. 5E, solid arrow). The gcm mutant had similar lch neuronal projection defects, where lch projections bifurcated just outside the CNS. Some lch neurons extended into the CNS along the anterior fascicle and the remainder migrated along a more posterior route (Fig. 5B, asterisk). Because the lch axons did not all migrate to the same location within the CNS as they would in the wild type, it follows that sensory axon misrouting at the TZ can cause incorrect migration within the CNS. Using Nomarski optics, the region where neurons stalled and began to exhibit pathfinding errors coincided with the ventralmost border of the abdominal muscles. In some cases, we observed sensory axons that had migrated to a much lower focal plane in the ventral CNS region than did those in wild types (Fig. 5F). In some embryos, when the ventral nerve cord was dissected off we observed lch axons projecting along the ventral muscles that lie just beneath the CNS. We also observed that in the absence of peripheral glia, lch neurons did not consistently migrate along VUM motor neuron tracts, which normally fasciculate with the lch neurons (Fig. 5D, arrow). Therefore, sensory neurons will not necessarily migrate along established motor axon routes, as demonstrated above. The data suggest that peripheral glia participate in guiding the entry of sensory axons into the CNS. Effects of Peripheral Glial Loss on Motor Axon Pathfinding We further investigated the potential role of peripheral glia as intermediate targets for the formation of motor neuron pathways. From anatomical analysis during early neural development, we observed that pioneer motor axons migrate through arrays of peripheral glia at the lateral edge of the CNS (Fig. 2) (Sepp et al., 2000). The migrational routes of the axons appear to accommodate the configuration of the peripheral glia in this region (Fig. 2C). In glial-ablated rq286::grim and gcm embryos, the pioneer motor axons exited the CNS along abnormal trajectories compared to the highly stereotyped pathways observed in wild types (Fig. 3F, solid arrow; Figs. 6A, 6C, and 6E, arrows). The motor axon phenotype of the repo::grim embryo was similar to that of the gcm mutant (data not shown). We observed evidence of motor neuron stalling, in that motor axons in some segments appeared delayed compared to others in different segments (Figs. 3F, 3H, 6C, and 6E). Repo-staining often obscures pioneer motor axon axons and peripheral glia cannot be accurately differentiated from other glia with the Repo marker; thus glial-ablated embryos could not be quantified for motor axon stalls at early stages. However, in gcm stage 13 and stage 14 embryos, stalling of motor axon pioneers was observed in 38% (n 86) of abdominal hemisegments, by comparing mutant axonal length to the wild type for the respective stages. Since motor axons stalled as they exited the CNS, it is likely that their growth cones experienced difficulties in migratory decision making. However, this initial difficulty in navigating the TZ could be overcome in both the glial-ablated rq286::grim and gcm, and in repo::grim embryos, given that motor neurons were capable of migrating to their appropriate muscle target regions in both cases (Figs. 6D, 6F, and 6H). The peripheral nerves were not entirely wild type in all mutants because they appeared defasciculated (Table 1B). In the gcm mutant, where many more CNS glia aside from the peripheral glia fail to differentiate, it is likely that many of the peripheral nerve defects are attributable to problems that arose during development inside the CNS. In support of this, the most severely affected segments of gcm mutants show concomitant CNS defects (Fig. 6F, arrows). The data suggest that peripheral glia may guide the initial migration of motor axons out of the CNS as prefiguring the TZ, although this function is not essential for later motor axon pathfinding within the PNS. Sensory Axon Pathfinding Relative to Motor Axons Sensory axons contact both peripheral glia and motor axons as they migrate to the CNS. Thus it is likely that both cell types play a role in guiding sensory axon migration. To gain a better understanding of the degree to which peripheral glia participate in sensory axon guidance to the CNS, we analyzed both motor and sensory axon phenotypes together in the gcm embryo by double antibody labeling of sensory and motor neurons for confocal microscopy. In this way, we could determine whether sensory axons rely on the presence of motor axons in their migration to the CNS. In wild-type embryos, sensory axons of the ventral (v) and ventral (v ) clusters fuse together to form the posterior fascicle (pf). To do this, the v cluster axons turn dorsally to meet with v cluster axons (arrowheads, Figs. 4A and 7A). The posterior fascicle fuses with the SNa motor axon root at approximately the same position where the v and v sensory branches meet and where the ISNb motor neuron branch leaves the ISN (Fig. 7A). In the gcm mutant we observed significant disruption of this stereotypic pattern of fasciculation. In all hemisegments analyzed (n 42), v cluster axons failed to carry out their early dorsal migrations (Figs. 3E, 7B 7D). As a result, they fasciculated with their v cluster partners at more ventral positions. This

13 Developmental Function of PNS Glia 59 phenotype was also observed in peripheral glial-ablated embryos (Fig. 4B). In some cases, the v cluster axons stalled completely [2.3% (n 42) (arrow, Fig. 7D)] or did not fasciculate with SNa motor axons at all [9.5% (n 42) (concave arrows, Figs. 7B and 7C)]. Both phenotypes are never observed in wild-type or control embryos. In wild-type embryos, the sensory axons of the anterior fascicle (af) fuse with the ISN motor branches in the ventral regions of the embryo. In the absence of glia, we observed sensory axons of the anterior fascicle (af) straying anteriorly from motor bundles of the ISN (solid arrow, Fig. 7C). The sensory axon pathfinding errors did not result from the absence of motor neurons. It is possible that the incorrect crossing of neurons at the CNS/PNS boundary causes further sensory axon pathfinding errors within the CNS, given that the CNS patterning of the axon bundles are variable from segment to segment compared to that of wild types (compare asterisks, Figs. 7A 7D). This confirms that peripheral glia play an important role in helping sensory axons migrate to and navigate their correct CNS entry points. DISCUSSION Previous anatomical studies of embryonic development in Drosophila suggested that peripheral glia may serve as guidepost cells for axon migration (Jacobs and Goodman, 1989; Klämbt and Goodman, 1991; Sepp et al., 2000). We assessed this hypothesis by analyzing embryonic nervous system development in the absence of peripheral glia. Our results demonstrate that peripheral glial cells are required for directing axon guidance across the CNS/PNS transition zone. The peripheral glia guide axons to the TZ, such that the major nerve roots form in close proximity to one another. During late stages of embryonic development, the peripheral glia are necessary for tight bundling and proper fasciculation of the peripheral nerves. Therefore, the peripheral glia have important roles in nervous system development throughout embryogenesis. Peripheral Glial Guidance of Motor Neurons From an anatomical perspective, peripheral glia could serve as very effective intermediate targets for pioneer motor axon pathfinding. Pioneer motor axons migrate directly toward the peripheral glia shortly after axogenesis begins. The pioneers appear to make stereotyped growth cone contact with the peripheral glia as they migrate peripherally (Auld, 1999; Sepp et al., 2000). The conformation of peripheral glia at the CNS/PNS boundary delineates the eventual migratory routes of the pioneers in the TZ region, as if the axons are funneled through the glial cone-shaped arrays. In both peripheral glial-ablated and gcm mutants, the stereotypical migrations of motor nerve pioneers across the TZ are disrupted. Therefore, the data suggest that by prefiguring the TZ, peripheral glia initially guide motor axons into the periphery. In wild-type embryos, once motor nerve pioneers reach the PNS, they migrate well in advance of extending peripheral glial processes (Sepp et al., 2000). As the pioneer motor axons reach the developing somatic muscle field, it is likely that peripheral glia no longer contribute to axon pathfinding decisions. In peripheral glial-ablated embryos, the errors motor nerve pioneers make as they cross the TZ can eventually be overcome. In the mature peripheral glialablated embryo, the most striking defect of motor neurons is defasciculation and aberrant separation of ISN and SN nerves, rather than incorrect nerve patterning, as was also observed in gcm mutant embryos (Jones et al., 1995; Hosoya et al., 1995; Vincent et al., 1996). Therefore, peripheral glia play only a minor role in the overall patterning of motor projections. An interesting comparison to the current study might be to previous experiments in the grasshopper, where the Segment Boundary Cell (SBC) glia were ablated, causing disruption of motor axon pathfinding out of the CNS (Bastiani and Goodman, 1986). However, the Drosophila equivalents of the SBCs are the ISN glia that, compared to the peripheral glia, are situated more proximally along the peripheral nerve root (Ito et al., 1995). The peripheral glial GAL4 line used in the current study is not expressed by the ISN glia; therefore, we could not determine the specific roles of the CNS glia that associate with the peripheral nerve root. In another study that analyzed the motor axon phenotype of the gcm mutant (Takizawa and Hotta, 2001), the initial motor axon projections in the gcm mutant within the CNS were found to be normal, and only later during pathfinding across the TZ were aberrations apparent, which included erratic distances separating the ISN and SN peripheral nerves, in agreement with our results. Further pathfinding within the periphery was found to be relatively normal, and motor axons could locate their appropriate muscle targets in the examples presented (Takizawa and Hotta, 2001). Interactions of Peripheral Glia with Sensory Neurons During embryonic development, peripheral glia physically interact with sensory axons significantly more than with motor axons. In regions of the embryo where sensory axons are located apart but in close proximity to motor axons, the peripheral glia typically migrate and extend their processes along the sensory tracts (Sepp et al., 2000). By the end of embryogenesis, the peripheral glia ensheathe the entire lengths of sensory fascicles but do not associate with the distalmost branches of motor tracts. Peripheral glia further extend their processes and entirely ensheathe motor nerves only during larval development (Sepp et al., 2000). Given the close anatomical associations with peripheral glia and sensory neurons, it is not surprising that when peripheral glia are ablated, more striking defects are observed with sensory rather than with motor neuronal development.