|Home | About | Journals | Submit | Contact Us | Français|
Glial cells play important roles in the developing brain during axon fasciculation, growth cone guidance, and neuron survival. In the Drosophila brain, three main classes of glia have been identified including surface, cortex, and neuropile glia. While surface glia ensheaths the brain and is involved in the formation of the blood-brain-barrier and the control of neuroblast proliferation, the range of functions for cortex and neuropile glia is less well understood. In this study, we use the nirvana2-GAL4 driver to visualize the association of cortex and neuropile glia with axon tracts formed by different brain lineages and selectively eliminate these glial populations via induced apoptosis. The larval central brain consists of approximately 100 lineages. Each lineage forms a cohesive axon bundle, the secondary axon tract (SAT). While entering and traversing the brain neuropile, SATs interact in a characteristic way with glial cells. Some SATs are completely invested with glial processes; others show no particular association with glia, and most fall somewhere in between these extremes. Our results demonstrate that the elimination of glia results in abnormalities in SAT fasciculation and trajectory. The most prevalent phenotype is truncation or misguidance of axon tracts, or abnormal fasciculation of tracts that normally form separate pathways. Importantly, the degree of glial association with a given lineage is positively correlated with the severity of the phenotype resulting from glial ablation. Previous studies have focused on the embryonic nerve cord or adult specific compartments to establish the role of glia. Our study provides, for the first time, an analysis of glial function in the brain during axon formation and growth in larval development.
The function of the brain relies on populations of neurons whose axon tracts form an invariant and highly specific network of connections. To unravel the mechanisms that control the formation of this axonal network is one of the key questions in developmental neurobiology. The Drosophila brain offers a favorable model for this kind of study. It is composed of a relatively small number of neural lineages, each lineage forming a genetic and developmental unit derived from a single neuroblast. Lineages also form structural units, in the sense that axons belonging to one lineage form a cohesive fascicle that interconnects specific domains (“compartments”) of the brain. In this paper we explore the relationship between axon fascicles and glia in the larval Drosophila brain, and ask in how far glia is involved in the growth and patterning of fascicles.
After delaminating from the neuroectoderm of the early embryo (Urbach et al., 2003; Younossi-Hartenstein et al., 1996), neuroblasts divide in a stem-cell like manner to produce approximately 10–20 primary neurons each. Neurons of one lineage emit axons that fasciculate and form the primary axon tracts (PATs) (Nassif et al., 1998; Younossi-Hartenstein et al., 2006). The PATs arborize in the late embryo to generate a stereotypical set of neuropile compartments which are later ensheathed by the cortex and neuropile glia (Hartenstein et al., 1998; Pereanu et al., 2005; Younossi-Hartenstein et al., 2003). During the larval stage neuroblasts re-initiate mitosis to produce the secondary neurons of the brain (Ito and Hotta, 1992). The secondary neurons also form cohesive bundles (secondary axons tracts, SATs) along the pre-established PATs. The SATs remain unbranched until pupal stages (Nassif et al., 2003; Pereanu and Hartenstein, 2006). In the pupal brain, SATs send dendritic and axonal arbors into the existing compartments of the larval brain; in addition, they establish a number of new, adult-specific compartments (Pereanu W., personal communication).
A wealth of experimental work has shown that axonal pathway choice depends on both intrinsic cues and extrinsic signals. Intrinsic cues are proteins (e.g., membrane receptors, cytoskeleton associated proteins) expressed by a given neuron in response to transcription factors inherited from the neuronal progenitor. Extrinsic cues are diffusible or local signals presented to the growing axon by cells that it encounters. During pathway guidance, two major sources of signals have been identified: pioneer neurons and glial cells. Pioneer neurons are early differentiating neurons whose cell bodies and axons are laid out in a scaffold along which the bulk of later born neurons fasciculate (Goodman CS, 1993; Nassif et al., 1998). Glial cells, which are born at around the same stage as pioneer neurons, also form segmentally iterated landmarks, or elongated tracks, that may be involved in axonal directional guidance. This has been studied in the Drosophila and grasshopper embryonic nerve cord (Jacobs and Goodman, 1989); for example, a subset of neuropile glia called “longitudinal glia” are required for the proper formation of connectives (Hidalgo and Booth, 2000; Hidalgo et al., 1995; Jacobs, 1993; Takizawa and Hotta, 2001); midline glia are required for commissure formation controlled, in part, by Neurexin IV and Wrapper-based adhesion (Stork et al., 2009; Wheeler et al., 2009) and robo/slit interactions (Kidd et al., 1999; Klambt and Goodman, 1991). In the embryonic peripheral nervous system, surface glia are required for the guidance of peripheral nerves (Gorczyca et al., 1994; Sepp et al., 2001) and embryonic motor neurons (Takizawa and Hotta, 2001). In the visual system, maturation of the lamina glia is required for correct R axon projection patterns (Yoshida et al., 2005).
Pioneer tracts and PATs of the Drosophila embryonic brain appear prior to the formation of glial sheaths, suggesting that glia may not be significantly involved as a guidance mechanism during this phase of development (Younossi-Hartenstein et al., 2006). This is very different from the situation when secondary axon tracts extend throughout the brain during the larval phase of development. At this stage, three major classes of glial cells have differentiated and are closely associated with the emerging secondary lineages and their axon tracts. Surface glia, consisting of large, squamous cells, surrounds the brain. Specialized intercellular junctions, called septate junctions, form the substrate of the blood-brain-barrier (Stork et al., 2008). Signals derived from surface glia have been implicated in the control of neuroblast proliferation (Ebens et al., 1993). Cortex glia is formed by highly branched cells that are intermingled with neuronal cell bodies in the brain cortex (Pereanu et al., 2005). Lamelliform processes of cortex glia extend in all directions away from the cell body and form sheaths around neuronal somata (trophospongium; (Hoyle et al., 1986; Pereanu et al., 2005). Apical processes of cortex glial cells also interact with surface glia and neuroblasts; basal processes form part of the thick glial lamella (“neuropile glial sheath”) that separates cortex from neuropile. The neuropile glial sheath incorporates another type of glia, called neuropile glia. Somata of neuropile glial cells are flattened and aligned at the cortex-neuropile glial boundary. Apart from contributing to the neuropile glial sheath, fine processes protrude into the neuropile where they form a dense glial reticulum that is in contact with dendrites, axons and synapses of the neuropile (Pereanu et al., 2005).
Several aspects of the relationship between glia and neural lineages of the larval brain have been addressed in previous studies (Dumstrei et al., 2003; Younossi-Hartenstein et al., 2006; Younossi-Hartenstein et al., 2003). It was shown that cortex glia forms chambers that enclose individual neuroblasts and their newly born progeny (ganglion mother cells, undifferentiated neurons). As neurons mature in deeper layers of the cortex, they are individually wrapped by the cortex glia. Cortex glia also envelops the secondary axon tracts in the cortex. Interactions between cortex glia and neurons, in part mediated by the DE-cadherin adhesion protein, are essential for the normal layering of neurons in the cortex (Dumstrei et al., 2003). When entering the neuropile, secondary axon tracts penetrate the neuropile glial sheath; glial envelopes or tracks also accompany SATs within the neuropile to various degree. However, the interaction between lineage associated axon tracts and neuropile glia within the larval brain has not yet been studied. We speculated that this interaction may play an essential role in axon tract formation. Thus, within the neuropile, SATs of neighboring lineages frequently converge and form larger fascicles; in other cases, tracts may come close to each other and cross, without fasciculation. It stands to reason that glial processes that do or do not isolate SATs could influence SAT pathway choices.
To address this question, we first mapped the association between SATs and glia in the third instar brain using confocal and transmission electron microscopy. We next selectively eliminated cortex and neuropile glia in the central brain and analyzed SAT morphology. We find that neuropile glia either wraps SATs, borders tract systems, or surround SATs as a fine reticulum. Simultaneous elimination of both cortex and neuropile glia leads to axon guidance and fasciculation phenotypes in the larval and adult brain, suggesting that glia of the brain, like those of the ventral nerve cord, is required for growth and guidance of the SATs.
The larval secondary neurons and axon tracts were labeled with an antibody against the Neurotactin protein (BP106; Developmental Studies Hybridoma Bank) and adult axon tracts were labeled with an antibody against the Neuroglian protein (BP104; Developmental Studies Hybridoma Bank). An antibody against the Drosophila N-Cadherin protein (DN-Ex#8, Developmental Studies Hybridoma Bank) was used to mark the neuropile in both larva and adult. Apoptotic cells were labeled with anti-cleaved Caspase-3 antibody (Cell Signaling Technology, Cat # 9661S). All glia were labeled with an antibody against the Repo protein (8D12; Developmental Studies Hybridoma Bank). Nirvana2-GAL4;UAS-GFP (Sun et al., 1999); Bloomington #6795) was used to drive protein expression in the cortex and neuropile glia. The adult DALcl1 lineage was visualized using the STAT92E10XGFP reporter line, a gift from Dr. GH Baeg (Bach et al., 2007). Apoptosis was induced using UAS-rpr (Bloomington #5823) as well as UAShid,UASrpr;;UASlacZ (Zhou et al., 1997). UASlacZ was used for controls. Conditional inhibition of UAS/GAL4 was accomplished via incorporation of the noc/Cyo;tubGAL80[ts] line (Bloomington #7018).
Antibodies were used at the following dilution: anti-Neurotactin 1:10, anti-Neuroglian 1:40, anti-DNCadherin 1:20, anti-cleaved Caspase-3 1:1000, and anti-Repo 1:10. Secondary antibodies were Alexa 546-congugated anti-mouse Ig 1:500 (Molecular Probes, A11030); Alexa 546-conjugated anti-rat Ig 1:500 (Molecular Probes A11081); Cy3-conjugated anti-rabbit Ig 1:200 (Jackson Immuno Research, 711-165-152); Cy5-conjugated anti-mouse Ig 1:100 (Jackson Immuno Research 115-175-166); and Cy5-conjucated anti-rat Ig 1:100 (Jackon Immuno Research, 112-175-102). For antibody labeling, standard procedures were followed (Ashburner, 1989). In brief, brains were dissected in 1XPBS pH 7.4 and fixed in 4% formaldehyde 30 min at room temperature or overnight at 4 degrees Celsius. Fixed brains were washed with 1X PBS at least 30 min and either stored in methanol at −20 degrees Celsius or further processed. After at least three 15 min washes in 1XPBT (tritonX .1–.3%), brains were incubated in 10% normal goat serum (Capralogics Inc, GS0500) for at least one hour at room temperature, followed by primary antibody incubation overnight at 4 degrees Celsius. Brains were then washed at least 30 minutes in 1XPBT three times, incubated again in 10% NGS for at least one hour, and left in secondary antibody for two hours at room temperature or overnight at 4 degrees Celsius. Finally, brains were washed at least 15 minutes, four times, in 1X PBT and mounted on glass slides in vectashield (Vector Laboratories, H-1000). Brains were viewed in the confocal microscope (20X or 40X objective; MRC 1024ES microscope with Radiance 2000 using Laser sharp 2000, version 5.2 build 824 software; Bio-Rad, Hercules, CA). Confocal sections were taken at 2 micrometer intervals for all preparations. We used Imgaej1.41d for image analysis and to generate merged confocal sections.
Larval brains were dissected in cold PBS and fixed in 4% paraformaldehyde, 2.5% glutaraldehyde in 0.1M phosphate buffer, pH 7.3, for 24 hours at 4°C. The brains were post-fixed for 60 minutes with 1% osmium tetroxyde in 0.1M phosphate buffer, pH7.3, at 4°C. Specimens were washed several times with distilled water and dehydrated in graded acetone series at 4°C. Specimens were incubated in 1:3, 2:2, 3:1 Epon:acetone ratio for 2 hours, 3 hours, and overnight, respectively. Brains were transferred to unpolymerized Epon and incubated overnight. They were then transferred to molds, oriented and placed at 60°C for 16 hours to permit polymerization of Epon. Blocks were sectioned at 0.06µm and mounted on slot grids and treated with uranyl acetate and lead citrate.
We generated either UAS-rpr; nrv2-Gal4,UAS-GFP/Cyo or UAShid,UASrpr;nrv2-GAL4,UAS-GFP/CyO;tubGal80ts larva for the experimental group and UASlacZ;nrv2GAL4,UAS-GFP/Cyo;tubGal80ts larva for the control group using the stocks listed above. For the initial experiments, embryos were laid and raised the entire time at 29 degrees Celsius on standard fly food. To initiate late-onset apoptosis in glia, tubGal80ts-containing embryos were laid at 18 degrees Celsius until wandering larval stage, at which point they were transferred to 29 degrees Celsius until dissection.
Flies were left to lay eggs overnight at 29°C on standard food. In duplicate or triplicate, ten first instar larva were manually picked from control and hid,rpr expressing progeny and placed on 60×15 mm standard food plates and returned to 29 degrees Celsius. Notes were recorded on growth and survival over the span of ten days. The experiment was repeating three times, and averages were taken.
We traced Neurotactin-positive tracts with the Labelfield module of Amira 3.1.1 (Mercury Computer Systems, Inc.). Labeled tracts were then three-dimensionally rendered and smoothed using the Amira SurfaceGen module.
To observe the association of lineages with the surrounding glia at third instar, we expressed GFP in cortex and neuropile glia using the nirvana2 (nrv2)-GAL4 line and visualized the lineages with an antibody against Neurotactin, a transmembrane glycoprotein that is highly enriched in the nervous system (Hortsch et al., 1990). Using this marker, all secondary lineages are individually recognized (Pereanu and Hartenstein, 2006).
We first investigated the wrapping of SATs by glia in the brain cortex. The cortex of a third instar brain, from deep to superficial layers, is composed of the embryonic-born (primary) cell bodies, early larval born (secondary) cell bodies, neuroblasts, cortex glia, and surface glia. In the deep layers of the cortex, the axons of the secondary neurons within each lineage fasciculate to form the SATs. The SATs are wrapped individually by cortex glia as they project toward the cortex/neuropile boundary (Fig. 1A). Even closely associated SATs are segregated by a thin glial septum, as visualized by confocal (Fig. 1E–G) and electron microscopy (Fig. 2A).
We next asked if the SATs remain segregated by glia in the neuropile. We observe three scenarios of SAT/glia interaction: (1) complete wrapping; (2) partial bordering (“glial tracks”); (3) no stereotyped association (see cartoon Fig. 1D).
To understand the role of glia during SAT guidance and morphology, we selectively eliminated the cortex and neuropile glia at a stage immediately prior to SAT formation. Towards this end, we ectopically expressed Hid and Reaper with nrv2-GAL4, a technique also utilized in previous studies (Pereanu et al., 2007a). This driver line is only expressed in neuropile and cortex glia, and not in surface glia (Fig. 3A, inset). Therefore, the blood-brain-barrier formed by surface glia (Pereanu et al., 2005; Stork et al., 2008) remains intact. This is essential for larval survival; simultaneous elimination of all glial types (Bernardoni et al., 1997; Pereanu et al., 2005) or the abnormal formation of surface glia alone (Bainton et al., 2005; Schwabe et al., 2005) results in embryonic lethality. It is also worth noting that primary lineages, which form before the onset of glial differentiation, are unaffected in nrv2-GAL4>UAS-hid,UAS-rpr (UAShid, rpr) larval brains. For the remainder of this study, the term “glia-less” refers to the elimination of larval cortex and neuropile glia, while surface glia remains unaffected throughout development (Fig. 3I repo-positive, GFP-negative, cleaved Caspase-negative cells).
In nrv2-GAL4,UAS-GFP embryos, GFP expression becomes strong in the late embryonic stages (Sun et al., 1999), a time when glial sheaths in the cortex and neuropile have differentiated. Correspondingly, UAS-hid,rpr induced apoptosis begins around hatching (Fig. 3E,I). Glia-less first instar larvae are overtly normal; no difference can be visualized by crawling speed or behavior. The dissected brain, however, shows a decrease in GFP-positive glial processes throughout the brain (Fig. 3E) as compared to their WT counterpart (Fig. 3A). By contrast, GFP-negative, Repo-positive surface glia are unaffected (Fig. 3I). An increased amount of cleaved Caspase, an indicator of apoptosis (Kumar and Doumanis, 2000), is seen in the nrv2-GAL4>UAS-hid,rpr first instar brain and nerve cords (Fig. 3F, arrows). By second instar, the glia-less larvae begin to crawl more slowly, appear uncoordinated, and become generally less active. At this stage, we see massive reduction of cortex and neuropile glia. Glial apoptosis can be quite variable; frequently, one hemisphere of the same brain retains considerably more glia than the other (Fig. 3G). By early third instar, control larvae crawl into the food; however, glia-less larva remain atop the food, are very inactive, and are much smaller in length and girth than their WT counterparts, most likely due to decreased feeding.
In general, the glia-less third instar larvae either die or remain in the larval stage longer than the control larva, which pupate by day 4 of development at 29 degrees Celsius. The length of time a glia-less larva will take to pupate can take up to 7 days at 29 degrees Celsius. The variable degree of glial ablation likely accounts for the variability in growth and survival; ultimately an average of fifty percent of glia-less progeny eclose (Fig. 3J). The glia-less adults most likely account for the least affected larvae, retaining a sufficient level of glia to progress through development to eclosion; however the adults are uncoordinated and very short-lived, surviving no longer than a few days.
The most prominent result of glial ablation is abnormalities in the pattern of secondary axon tracts. Phenotypes can be classified as misguidance, truncation, or absence of the SAT. Under “absent” SATs we include cases where axons are truncated already in the cortex, or where misguidance of the SAT and its neighbors is so severe that we cannot determine its identity. The percentage of effected tracts among SATs analyzed range from 0% to 94% (Fig 4, lower graph).
If SATs are ranked according to the amount of glia associated with them in the neuropile (see previous section), this ranking correlates positively with the percentage of phenotypes seen per SAT (Fig. 4, upper graph). In other words: SATs that normally are associated with a lot of glia are generally affected more frequently, and show more severe phenotypes, than SATs that normally do not have much glial contact. For the ranking we score glia-SAT association in the following manner: A score of 5 indicates that an SAT is completely wrapped by glia; 4: SATs that extend along glial tracks or sheets; 3: SATs that are wrapped by glia over a short segment of their trajectory; 2: SATs are contacted by glial track over a short segment of SAT trajectory; 1: no overt association with glia (Fig. 4, cartoon). Table 1 summarizes SAT morphologies, trajectories, and scores their association with neuropile glia.
The SATs most strongly affected were those of the four mushroom body lineages that form the peduncle and lobes of the mushroom body. The mushroom body SATs join together at the cortex neuropile boundary (see Fig 1A). Once in the neuropile, the four SATs merge to form the peduncle; neuropile glia completely surround the peduncle, but do not project processes between the mushroom body SATs. The peduncle travels in a straight posterior to anterior direction. Near the anterior end of the neuropile, the peduncle branches into a dorsal and medial lobe. In glia-less brains, we find three distinct abnormal mushroom body morphologies: 1) mushroom body SAT splaying/misguidance, 2) mushroom body and non-mushroom body-SAT merging, and 3) premature (intracortical) mushroom body SAT fasciculation (Fig 5).
The variability in mushroom body phenotypes is consistent with the variability in glia elimination in general. (1) The mushroom body SAT splaying is characterized by random de-fasciculation and arbitrary trajectories away from the peduncle (Fig. 5I–L). (2) In addition to keeping the mushroom body SATs bundled and directing their straight forward trajectory, glia insulate these SATs from adjacent axon tracts. In glia-less brains we observe axons from the CP SATs, that normally cross over the peduncle in front of the calyx, de-fasciculate from their own SATs and join the peduncle (Fig. 5E–H). Mushroom body and non-mushroom body SAT fusion are never observed in control samples (Fig 5A–D), where a glial sheath segregates the mushroom body SATs from the CP SATs (Fig. 5B). (3) We frequently observed premature bundling of the mushroom body SATs in the brain cortex (Fig. 5I,O). Normally, the four lineages separately project into the neuropile, forming four tracts surrounding a central sphere of dendritic arbors, the calyx. In the glia-less brain, the lineages merge too soon in the cortex, frequently forming a misshapen, superficial, calyx (Fig. 5I,P).
To investigate the effect of glia ablation on other lineages, we selected representative SATs of the nine families of lineages and studied their pattern in 16 glia-less hemispheres of third instar brains. The lineages selected had SATs that were easily identified in wild-type brains by trajectories that stood out (i.e. they are non-linear, members of larger tract systems, or cross the midline). The lineages selected are listed in Table 1. These lineages can be grouped into four categories: 1) medial with high glial association (BAmas1/2, DAMd1, DAMv2/3, CP1); 2) medial with lower glial association (BAmv2); 3) lateral with high glial association (DALcm1, DALcl1, BLVa1-3); and finally 4) lateral with a low glial association (BAla3, DPMl1, DPLal1-3, BLD1-4). In the following we will consider the effect of glia loss on these categories of lineages.
Most SATs that cross the midline are completely wrapped by glia (score of 5) and, similar to the mushroom body SATs, show frequent abnormalities in glia-less brains. Of the medial glia-rich lineages, the CP1 SAT in particular shows a dramatic phenotype in which its medial projecting tract makes a sharp turn ventral at the midline rather than crossing to the contralateral lobe (37.5%, Fig. 6A,A” green box, Fig. S2A–C). CP1 also has a branch that extends anteriorly and ventral to the peduncle, and 18% of the ventral tracts showed misguidance/fasciculation abnormalities (data not shown). In other medial glia-rich lineages, we observe missing tracts (Fig. S2A, yellow arrows indicate DAMd tracts missing in right hemisphere), and abnormal morphology of SATs in the glia-less midline (Fig. 6A,A” green box) as compared to the control brain (Fig. 6B,B” green box). Lineages with a lower glial association along the majority of the distal SAT, such as BAmv2, show a smaller percentage of abnormal projections. In 25% of the BAmv2 tracts, its two projections fail to separate upon turning dorsal in glia-less brains (Fig. 6C,C” yellow box) compared to control brains (Fig. 6D,D” yellow box).
Based on the observation that midline-projecting larval SATs exhibit a high level of abnormal projections, we next ask if midline-projecting fascicles of the adult brain also show aberrant trajectories. Axon fascicles in the adult brain are visualized with anti-Neuroglian. In most cases, SATs of several lineages merge during pupal stages. It is therefore difficult without specific markers to verify the structural abnormalities of SATs diagnosed for the glia-less larval brains. In regard to the medial lineages, adult defects can be inferred. Several of the fascicles to which SATs of the medially projecting lineages contribute are abnormal, typically missing or truncated on one side (Fig. 7B, arrows) as compared to a control adult brain (Fig. 7A).
Third instar lateral lineages with a high glial association studied include BLVa1-3, DALcm1, and DALcl1. With the exception of BLVa1-3, which is rarely affected (SAT truncation in 6%), all other lateral lineages with glia-rich SATs show a high degree of abnormalities in glia-less brains. In the DALcm1, 31% of tracts are truncated at the point where DALcm1 meets the BAmd1 SAT (Fig. 6A,A’ red box). The remaining percentage of DALcm1 phenotypes (each only observed in one hemisphere) include loss of axon fasciculation and disorganized projection of the medial and ventral DALcm1 tracts (data not shown).
To study DALcl1 SAT morphology, we utilize the STAT92EGFP reporter line (Bach et al., 2007). This reporter is expressed in a large subset of DALcl1 neurons in the adult brain. In control adult brains, the SAT of STATGFP-positive neurons projects dorsal to the mushroom body peduncle (Fig. 8A,A”). In glia-less brains, 60% of the hemispheres show the STAT92EGFP positive tract as a projection ventral to the mushroom body (Fig. 8B,B”). We next looked at the terminal arbors of the tract. The DALcl1 axonal terminals form large, conspicuous terminal boutons in the lateral bulb, a small compartment flanking the central complex (Fig. 8A’). In DALcl1 axon tracts of glia-less brains, we observe ectopic terminal boutons directly ventral to the mushroom body spur, much more laterally than in wild type (Fig. 8B’, arrowhead). However, the majority of the tract terminates in a similar area to the control brain even after extending along a different path.
While most of the lateral SATs highly associated with glia have frequent misguidance phenotypes, lateral SATs with a low glial score exhibit defects at a lower rate, and these defects typically consisted in truncation/absence of the SAT. For example, select lateral glial-poor SATs normally enter the neuropile at the lateral edge and turn dorsally into the CPLd compartment (Fig. S2D left hemisphere, white arrow). However, these particular SATs are severely reduced in 25% of the glia-less brains, and in one brain a lump of truncated SATs at the cortex/neuropile boundary is identified (Fig. S2D, right hemisphere, white arrow). Other lateral glial-poor SATs enter at the lateral neuropile and project postero-ventrally toward the lateral face of the peduncle. Figure 6C,C’ (pink box) shows an ectopic trajectory along the lateral boundary of the neuropile from a set of lateral lineages. This trajectory is not evident in any control preparations (Fig. 6D,D’ pink box). Of the remaining lateral glial-poor lineages studied, only one brain contained a missing SAT (Fig. S2D, yellow arrows), and some of the lateral glial-poor lineages failed to show any aberrant phenotypes.
Axonal and dendritic arborizations of lineage fascicles form the compartments of the larval and adult brain. First, in the late embryo and early larva, terminal arbors (both axonal and dendritic) sprouting from primary axon tracts (PATs) form the larval compartments. Most of these persist through metamorphosis into the adult brain; typically, they merely grow and, in some cases, are further subdivided into distinct sub-compartments. Thus, during metamorphosis, secondary axon tracts produce their own terminal arbors. These arbors grow into the pre-existing compartments of the larva. In addition, new, adult-specific compartments are generated. These include the central complex (fan-shaped body FSB, ellipsoid body EB, noduli), the protocerebral bridge, the lateral bulbs (input region of central complex), the antennal lobe, the antenno-mechanosensory and motor center (AMMC), and the optic tubercle.
If glia is removed during larval stages, as in the experiments reported here, one would expect abnormalities to be more prominent in compartments formed de novo from secondary lineages. This turns out to be true for some compartments but not others. For example, in glia-less adults, 73% (n=15) of cases have pronounced defects in the morphology of the EB. Typically, the EB is flattened in the dorso-ventral axis and is not closed ventrally in the absence of cortex and neuropile glia (Fig. 9D–F) as compared to a control preparation (Fig. 9A–C). The next most-affected adult-specific compartment includes the AMMC, in which 25% (n=16) are severely reduced in size if not completely missing. In contrast, the FSB, protocerebral bridge, and optic tubercle show no severe morphological defects (data not shown), although the detailed neural layering and connectivity was not tested; therefore we can not rule-out a functional change in these compartments.
The final adult-specific compartment we analyze is the antennal lobe (AL), which is formed during pupal development in place of the degenerated larval BA compartment (Jefferis et al., 2004). In the AL, olfactory neuron terminal arbors organize into glomerular structures and synapse with the dendrites of uni- and multi-glomerular projection neurons, which then project to the mushroom body calyx and the lateral horn (Lai et al., 2008). AL glia elimination in the moth Manduca sexta results in disorganized glomerular architecture (Rossler et al., 1999); however it has been hypothesized that glia do not play a major role in Drosophila AL organization due to a late onset of glia wrapping, after glomeruli have already formed (Jefferis et al., 2004; Oland et al., 2008). Normally, the AL has a well-described architecture with readily identifiable glomeruli that have been carefully mapped and named (Laissue et al., 1999). When we eliminate glia at third instar, 43% (n=14) of the ALs are smaller in size and lack clear glomerular boundaries (Fig. S3D–F) as compared to the control brains (Fig. S3A–C). In the most severe case, the glomeruli aren’t identifiable at all (Fig. S3D–F, right AL). Therefore, cortex and neuropile glia elimination in the larva at least partially disrupts the organization of the adult AL.
We next investigated compartment morphology when cortex and neuropile glia are eliminated after SAT projection. To this end, we utilized a temperature sensitive GAL80 transgene to conditionally express Hid and Rpr only after secondary lineages are born and project axons into the neuropile. Larvae were incubated at 18 degrees Celsius, the temperature at which GAL80ts is active, until late third instar, allowing secondary lineages to form in the presence of glia. At late third instar, the larvae were shifted to 29 degrees Celsius, rendering GAL80ts inactive, to induce glial apoptosis. While late onset of glial death still results in glia elimination in the adult (Fig 9H), neuropile compartments show a normal morphology (Fig. 9G–I). This suggests that the compartment abnormalities caused by glial loss stem from events that occur during larval and pupal development, such as secondary axon tract formation and branching.
Drosophila has been extensively used to analyze the developmental roles of glia in nervous system growth and maintenance. Most studies have focused on the embryonic nerve cord, due to its highly organized and relatively simple structure, in an attempt to understand the complex interactions between neurons and glia. Less attention has been given to the more complex larval brain, stemming in part from a lack of morphological characterization of the brain’s structure. Only recently did we develop a map that defines the position and axonal trajectory of each neural lineage within the larval brain (Pereanu and Hartenstein, 2006), providing an anatomical body of reference to which abnormal phenotypes resulting from genetic manipulations can be compared. In the present study, we have addressed the role of glia in larval brain development. Specifically, we establish the degree to which each lineage-related axon tract is associated with glia, to then ask how the pattern of SATs changes in the absence of the cortex and neuropile glia.
Secondary neurons, which are born during the larval period, form SATs that have to extend over relatively long distances, finding their way amidst a complex array of (primary) axons, dendrites, and glia. The association of glia and SATs varies for different lineages. SATs either 1) remained wrapped within the neuropile or joined other tracts that were then ensheathed as a larger tract system, 2) encountered strands of glial condensations, or 3) had no association with glial sheaths. The association between individual SATs and glia was highly invariant. Thus, if SAT A joined SAT B to form a larger tract system that became wrapped by glia, the same densities of glia could be observed in other brains for the corresponding SATs.
To address the role of glia during SAT fasciculation, growth, and guidance, we selectively eliminate cortex and neuropile glia, the two glial types in contact with growing SATs. Expression of the pro-apoptotic proteins Hid and Rpr were effective in inducing apoptosis of most cortex and neuropile glia by the early or mid larval stage. We can assume that the primary axon tract formation is not disrupted, given that expression of the nrv2-GAL4 driver line does not set in prior to stage 12 (Sun et al., 1999), PAT patterning is complete before glia invade the neuropile (Younossi-Hartenstein et al., 2006), and the first signs of apoptosis appear after hatching. In addition, the surface glia remain intact, providing general ensheathment around the brain, a functional blood-brain barrier, and potential signaling molecules used to control neuroblast proliferation.
Upon the elimination of glia, we see frequent abnormalities in the pattern of SATs. Importantly, the strength of an SAT phenotype appears to correlate with the degree of glial association of that SAT in a wild-type brain. Of all SATs analyzed, the mushroom body (associated with a complete glial sheath) exhibits the most severe defects, including complete SAT misguidance and aberrant fasciculation of neurites from adjacent SATs. In contrast, SATs that were less endowed with glia in the wild type typically had a normal projection pattern.
Our study does not differentiate between a role for glia in producing chemo-attractant signals or acting as a physical scaffold for guidance. In previous studies, ectopic expression of dominant-negative Drosophila E-cadherin in either cortex/neuropile glia or SATs themselves results in non-radial trajectories of SATs into the neuropile (Dumstrei et al., 2003). This suggests a requirement for SAT-glia adhesion as secondary axons project toward the cortex-neuropile boundary. However, the direction of neuroblast division is also disrupted with DE-cadherin knock-down, thus aberrant SAT trajectories may be a secondary effect of abnormal cell body layering within the cortex. In support of a signaling role for glia, the midline guidance defect of the CP1 SAT is reminiscent of the robo-slit phenotypes in the Drosophila ventral nerve cord (Kidd et al., 1999). Whether robo/slit signaling is also used between SATs and glia in the brain is a question that warrants further investigation.
The situation that SATs in part require glia for pathway guidance in the neuropile is different than the embryonic brain in which pioneer axons and PATs are formed in the absence of extensive glial processes. Why are SATs different? One can imagine the scenario in which the PATs generate the neuropile de novo, forcing cell body movement outwards as the central neuropile grows. By first instar, a full neuropile is established, and SATs must guide through a dense maze of neurites, glia, and trachea. It therefore appears that: 1) PATs establish the initial connectivity of the brain, 2) glia grow in around this initial scaffolding, and finally 3) SATs use both the PAT scaffolding and the glial boundaries for guidance into and around the neuropile.
Insects have long been used to evaluate glial-neuronal interactions from embryonic to adult stages. An important focus of these studies was whether or not glia or axon tracts appear first, and in how far axonal pathfinding is disrupted if glia is ablated. In this regard, clear developmental differences have been found between distinct regions of the nervous system, as well as between insect species for homologous nervous system domains.
In the Drosophila ventral nerve cord, two subpopulations of neuropile glia were studied in the context of axonal pathfinding: midline glia and longitudinal glia (reviewed in (Klambt et al., 1996). Both types of glial cells appear around the same stage when pioneer neurons extend their axons. The proper number and positioning of midline glia is clearly required for the formation of commissural axon tracts (Sonnenfeld and Jacobs, 1995). The loss of longitudinal glia (by ablation and in embryos mutant for the gcm gene) primarly affects the defasiculation and fasciculation events of longitudinal pioneer tracts, subsequently affecting the follower neuron trajectories (Hidalgo and Booth, 2000). Note that gcm is required for all classes of glia, including surface glia; thus, in very late gcm mutant embryos, severe disruptions of the entire neuropile result from the fact that, with the onset of embryonic movement, the CNS lacking surface glia is literally “shredded to pieces” (Pereanu et al., 2007b).
In the embryonic Drosophila peripheral nervous system, ablation of the peripheral glia (i.e. exit glia) via targeted overexpression of the cell death genes grim and ced-3 lead to aberrant pathfinding of both motor and sensory axons as they exited the CNS; although the motor neurons eventually overcame the absence of peripheral glia finding correct muscle targets, suggesting a limited role for the peripheral glia in the initial trajectory of motor neurons (Sepp et al., 2001). In grasshopper, ablation of the cell-segment boundary glial guidepost cell lead to more severely aberrant axon trajectories (Bastiani and Goodman, 1986). Ablation of glia surrounding the antennal lobe of adult Manduca generates olfactory axon defasciculation and misguidance (Rossler et al., 1999). Finally, guidance phenotypes have also been observed with disruption of the Drosophila lamina glia (Tayler and Garrity, 2003; Yoshida et al., 2005); R1–R6 photoreceptor axons show aberrant guidance past the lamina into the medulla of the optic lobe (Tayler and Garrity, 2003).
An important aspect of the developmental role of glia added by our study is the focus on the correlation between closeness of axon tract/glia association, and axon tract abnormality in the absence of glia. In other words: the nervous system is formed by a large number of fascicles, and these fascicles vary in their degree of glial wrapping. It would be misleading when carrying out a genetic study to only focus on a single fascicle (or small subset of fascicles), and extrapolate from the phenotype observed for this fascicle onto the brain as a whole. In our analysis, SATs of several lineages, notably those that in normal brains have little glia covering them in the neuropile, show few abnormalities in glia-less brains. By contrast, other lineages were affected in the majority of cases, and typically, these lineages also were the ones whose SATs were associated more closely with glia.
Neuropile compartments are formed by terminal branches of axons and dendrites and their synapses. For example, the antennal lobe of insects consists of the axonal terminals of sensory neurons located in the antenna, and dendritic terminals of antennal projection neurons and local interneurons located in the deuterocerebrum, aside from a relatively small number of other modulatory neurons. Sensory neurons expressing the same olfactory receptor all converge onto the dendrites of a small number of projection neurons to form an olfactory glomerulus (Laissue and Vosshall, 2008). In many insects, notably Manduca, olfactory glomeruli are individually compartmentalized by neuropile glia, and it has been shown that glia plays a prominent role in establishing the glomeruli organization. According to the prevailing view, glomeruli are initially ordered by the specialized endings of sensory terminals into protoglomeruli; however, glial processes soon invade the space in between protoglomeruli and restrict the arborization of receptor axons (Baumann et al., 1996). Therefore, in Manduca antennal lobes, glia is required for early maintenance of the glomerular map.
A recent analysis of the time course of glial development in the Drosophila antennal lobe (Oland et al., 2008) suggested that in this species, glia plays an even lesser role, since glomeruli are only incompletely, and at a late time point, wrapped by glia (Jefferis et al., 2004; Oland et al., 2008). While glia were never ablated in those studies, later work found that elimination of Neuroglian from midline glia resulted in an inability of ORN axons to cross through the antennal commissural tract to the contralateral lobe (Chen and Hing, 2008). In our study, most adult brains lacking cortex and neuropile glia still form antennal lobes with glomeruli, however in some samples the glomeruli are poorly defined, and cannot be identified by their position and shape in the antennal lobe. The variability in phenotype penetrance likely stems from larvae containing the most glia surviving to eclosion; therefore, a relatively normal looking adult brain could stem from a weak glial apoptosis early in development. Whether the antennal lobe disorganization is a secondary defect due to a lack of definition normally provided by the presence of glia, a result of SATs that normally contribute to the AL misguiding or truncating early, or a maintenance defect in which axons begin inappropriately intermingling among glomeruli is not clear. Perhaps the Drosophila antennal lobe glia is required for glomeruli maintenance even after glomeruli organization is established, and future studies will hopefully address this possibility.
In the post-embryonic midline of Drosophila, ablation of the transient interhemispheric fibrous ring (TIFR), a transient population of midline glia, by ectopic expression of the pro-death gene hid, generates defects in the adult central complex (Hitier et al., 2000). It is unclear, however, the cellular event that is effected to cause the defects. Our study suggests that morphological abnormalities in adult compartments from glial manipulation are due to the misguidance of larval SATs to the correct neuropile compartment in the brain, affecting the formation of adult neuropile structures. Here, we present evidence that glia is an important mediator of axon guidance in the Drosophila larval brain; the mechanism for glia-neuron communication during this process is an exciting area for future investigation.
We are grateful to the Bloomington Stock Center and Dr. GH Baeg for providing us with fly stocks. This work was supported by NIH Grant RO1 NS29357-15 to V.H., the Ruth L. Kirschstein National Research Service Award GM007185 to S.S., and IMSD grant GM55052 to I.O.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.