Synaptogenesis in the CNS has received far less attention than the development of neuromuscular synapses, although only central synapses allow the study of neuronal postsynaptic mechanisms and display a greater variety of structural and functional features. This neglect is attributable mainly to the enormous complexity of the CNS, which makes the visualization of individual synapses on defined neuronal processes very difficult. We overcome this obstacle and demonstrate by confocal microscopy the specific arrangement of output synapses on individual neurites. These studies are performed via genetic mosaic strategies in the CNS of the fruitfly Drosophila melanogaster. First, we use targeted expression of synaptic proteins by the UAS/Gal4 system. Second, we apply a newly developed transplantation-based mosaic strategy that takes advantage of the intrinsic regulation and localization of synaptic proteins in single-cell clones. We propose the existence of three distinct neuritic compartments: (1) primary neurites that appear to form the main transport pathways and are mostly void of output synapses, (2) neuritic compartments that contain output synapses, and (3) neuritic compartments that are postsynaptic in nature. In addition we show that mutations of the kakapogene have no obvious effect on the distribution of output synapses in the CNS, whereas neuromuscular synapses are severely reduced. This suggests that synaptogenic mechanisms in the CNS might differ from those at neuromuscular junctions.
During development of the nervous system, synapses of the appropriate type have to differentiate and be positioned correctly, so that information flow and processing are orchestrated into functional neuronal circuits. Mechanisms underlying the complex process of synapse formation are poorly understood. To address such mechanisms we chose the embryonic nervous system ofDrosophila melanogaster because of its defined cellular and developmental context in combination with its genetic amenability and the availability of numerous molecular tools (for review, see Budnik and Gramates, 1999). Like other insects (for review, see Burrows, 1996), Drosophila contains identified neurons; i.e., individual neurons can be recognized with respect to the shape and position of their somata and neurites, and many neurons display specific patterns of gene expression (Thomas et al., 1984; Doe and Technau, 1993). From larger insects we know that these identified neurons interact in stereotypic ways with other neurons; i.e., neuronal circuits can be traced back to their individual cellular elements (for review, see Burrows, 1996). So far, studies on synapse formation and function during embryogenesis have been focused mainly on neuromuscular junctions because of their experimental amenability (for review, seeBudnik and Gramates, 1999). However, the mechanisms that can be studied at Drosophila neuromuscular junctions are limited. First, the postsynaptic cells are muscles; thus neuronal postsynaptic mechanisms cannot be studied (e.g., postsynaptic densities are far more prominent at many synapses in the embryonic CNS than at neuromuscular synapses) (Prokop, 1999). Second, in addition to few neuropeptides and other potential neuromodulators, glutamate seems to be the main transmitter at neuromuscular junctions (NMJs) (Johansen et al., 1989), whereas a larger diversity of neurotransmitters are found to operate in the CNS (for review, see Prokop, 1999). Thus, potential mechanisms specific to nonglutamatergic synapses can be studied only in the CNS.
In the CNS of insects, neurons are monopolar, sending one primary neurite from the cortex (cell body region) into the neuropile (soma-free synaptic area) where they branch and form axons and neuritic arborizations with presynaptic and postsynaptic sites in yet unknown distributions. Ultrastructural analyses of neurons in larger insects have shown that neurites can be specialized (i.e., exclusively presynaptic or postsynaptic), or they can be of mixed nature (for review, see Burrows, 1996). Whether the same organizational principle also applies to the dramatically smaller Drosophila neurons, and more specifically, which neurites of various identified interneurons contain presynaptic or postsynaptic sites, has not been investigated so far.
Here we have addressed these shortfalls and show for the first time that output synapses are restricted in a reproducible manner to specific neuritic compartments of individual neurons in the trunk neuropile of Drosophila embryos. To this end we applied light-microscopic analyses in combination with mosaic techniques based on targeted gene expression and a newly developed strategy using established cell transplantation methods. These new insights and techniques are an important prerequisite for future work on synapse formation and structure in the Drosophila CNS.
MATERIALS AND METHODS
Fly stocks. Fly strains used in our studies are different mutant alleles of kakapo(kakaposf20 ,kakapoHG25 , short stop3 ) (vanVactor et al., 1993; Prokop et al., 1998), tubP-Gal4 (by courtesy of L. Luo) (Lee and Luo, 1999), elav-Gal4C155 (by courtesy of the Bloomington stock center) (Luo et al., 1994), Uas-mCD8-green fluorescent protein (GFP) (second and third chromosome; by courtesy of L. Luo and Bloomington stock center) (Lee and Luo, 1999),apterous-Gal4 (ap-Gal4; by courtesy of John Thomas) (Calleja et al., 1996),Uas-Synapsin-112 (this paper), Df(3)Synapsin97(Syn97 ) (Godenschwege et al., 2000),Uas-Synaptotagmin-hemagglutinin (Uas-syt-HA; by courtesy of I. Robinson) (Robinson et al., 2002),SynaptotagminAD4 (SytAD4 ) (Littleton et al., 1994), andUas-neuronal-Synaptobrevin-GFP13.6 (Uas-n-Syb-GFP; by courtesy of M. Ramaswami) (Ito et al., 1998). To obtain recombinant kakapo donor stocks (see Fig.5 G), tubP-Gal4 constructs were jumped from the third to the second chromosome [according to Robertson et al. (1988)]. Recombinations followed standard procedures (Greenspan, 1997).
Generation of Uas-Syn-112 transgenic flies. An Xho-PshAI 3.1 kb genomic fragment (containing the exon 1, intron 1, and part of exon 2 of theSynapsin gene) of the plasmid rescue vector of the SynP1 (Godenschwege et al., 2000) and a PshAI-XbaI 3.3 kb fragment of the Syn-1 cDNA (containing part of exon 2 and exons 3–13 of theSynapsin gene) (Klagges et al., 1996) were cloned into a Xho-XbaI-digested pP(UAST) vector (Brand and Perrimon, 1993). Transformation of Syn97 mutant flies with the vector containing the UAS-Syn-1 gene was performed as described elsewhere (Spradling and Rubin, 1982) and resulted in a fly stock carrying a third chromosomal insertion (UAS-Syn-112,Syn97 ).
Immunohistochemistry. Antibodies used in these studies were raised against presynaptic proteins Synapsin (mouse, 1:10) (Klagges et al., 1996), Synaptotagmin (courtesy of H. Bellen) (rabbit, 1:1000) (Littleton et al., 1993), and Cysteine string protein (courtesy of K. Zinsmeier) (Zinsmaier et al., 1990), the transmembrane domain protein CD8 (rat, 1:10; Caltag Laboratories, Burlingame, CA), and the epitope tag hemagglutinin (rat, 1:100; Boehringer Mannheim, Mannheim, Germany).
Dissection, fixation, and staining procedures followed standard protocols (Broadie, 2000). In brief, forceps were used to dissect out CNS from old embryos or larvae. All dissections were performed in external bath solution for electrophysiology (Broadie, 2000). After 30–60 min fixation in 4% paraformaldehyde in 0.05 mphosphate buffer, pH 7–7.2, tissues were washed for 1 hr in PBS with 0.1% Triton X-100 (PBT). Incubation with antibodies was performed in PBT without any blocking reagents, followed by incubation with commercial secondary antibodies (Jackson ImmunoResearch, West Grove, PA; dilution 1:200) coupled to fluorescent dye or horseradish peroxidase. Fluorescent analyses were performed on a Leica confocal microscope true confocal scanner SP2 and Leica confocal software. Horseradish peroxidase was detected via a color reaction in 0.5 mg/ml diaminobenzidine with 0.03% H2O2.
Cell transplantations. Cell transplantations were performed as described elsewhere (Prokop and Technau, 1993). In brief, donor and host embryos were of the same stage at ∼10 min after the onset of gastrulation. With use of a pulled and ground glass capillary, with an inner diameter of ∼10 μm, 10–20 cells were removed from the ventral neurogenic region of the donors at ∼0–30% ventrodorsal diameter and at 30% egg length, i.e., in the abdominal region [according to the early gastrula fate map (Prokop and Technau, 1993)]. Single precursors were carefully injected into the same area of the blastoderm of host embryos. Donor embryos were genetically labeled with the cell surface marker CD8, host embryos were mutant for certain synaptic proteins (genotypes described in Results; see Figs.1 B, 4 A). After transplantation, the host embryos were allowed to develop into the final embryonic stage (trachea filled with air). During this time the implanted precursor cells give rise to fully differentiated cell lineages. Most of the development occurred overnight at 25°C, and during the last few hours embryos were shifted to 29°C to enhance Uas-CD8 expression. Fully developed host embryos were dissected and immunostained as described above.
Quantitative analyses. To estimate boutons and output synapses of cell lineages, areas of the lineages were chosen in which synapses were not clustered too densely (see Table 1 legend). With use of Leica confocal software, serial confocal stacks were analyzed section by section, and Synapsin spots of this area were counted on the computer screen. Using Adobe Photoshop software, the circumference of the same area was drawn with the selection tool, and pixel content was determined with the histogram function. To normalize data, the pixel content was divided by the pixel number obtained by measuring and squaring the neuropile width of each respective specimen.
The synaptic area of Drosophila is segmentally organized
In the Drosophila ventral nerve cord (abdominal and thoracic part of the CNS), segmental ganglia are fused, although segmental units (neuromeres) can still be recognized by distinct landmarks such as the segmental nerves. Each neuromere contains an almost identical set of neuronal cell bodies in typical positions, each of which sends one process into a soma-free zone called neuropile. The neuropile is composed of two connectives (longitudinal tracts) that are connected across the midline in each neuromere by two commissures (transverse tracts) (Fig.1 B′). Within the neuropile, neurites of individual neurons are stereotypic and reproducible with respect to their shapes and positions (Udolph et al., 1993; Landgraf et al., 1997; Schrader and Merritt, 2000). All synaptic contacts of the CNS are restricted to neurites in the neuropile as revealed by antibodies against presynaptic proteins such as Synaptotagmin and Synapsin (Fig.1 A,A′). Their stainings show dense arrangements of little dots in the neuropile representing output synapses (i.e., presynaptic sites or zones). Within this almost homogeneous accumulation of output synapses, reproducible segmental pattern elements like transverse and circular gaps can be seen always in the same positions (Fig. 1 A,A′,arrows and arrowheads, respectively). Such gaps suggest that certain neurites in reproducible areas of the neuropile might be void of presynaptic sites. We investigated this possibility in larger detail and analyzed distributions of output synapses along individual neurites. To this end we used mosaic analyses based on targeted gene expression and cell transplantations.
Using targeted expression of presynaptic proteins to visualize output synapses
To monitor the localization of presynaptic proteins in small numbers of identified neurons, we first used the Gal4/Uas system of targeted gene expression (Brand and Perrimon, 1993) (Fig.2 A). We restricted our analyses to interneurons because they should form output synapses in the CNS (which might not be the case for insect motoneurons) (for review, see Burrows, 1996). To our knowledge, ap-Gal4 is the only Gal4 driver line with strong and reliable expression in a restricted number of interneurons of late embryos and larvae (two to three interneurons on either side of each neuromere in the ventral nerve cord) (Lundgren et al., 1995) (Fig.2 B–F′′′). First, we crossedap-Gal4 flies to Uas-CD8-GFP flies, carrying a construct of the cell surface marker CD8-GFP coupled to the Gal4-responding Uas-sequence. In their offspring (ap:: CD8-GFP animals) ap-Gal4 neurons can be visualized with anti-CD8 antibodies (Fig.2 B,E). Each cell body (Fig. 2,S) sends a neurite through the cortex (Fig.2 C,E,F, open arrows) that enters the neuropile on its lateral side. From the neuropile entry point, the primary neurites project transversely (Fig.2, T) toward the midline where they join a compact median longitudinal fascicle (Fig. 2, L). In the late embryo, small side branches can be seen coming off the transverse primary neurite (Fig 2 B, white arrow). These side branches are far more elaborate in the late larval nerve cord (Fig. 2 E, white arrow).
To visualize potential presynaptic sites (output synapses) ofap-Gal4 neurons, we crossed ap-Gal4 flies with different transgenic fly strains carrying Uas-constructs coding for presynaptic proteins (Fig. 2 A). To this end, we usedUas-neuronal-Synaptobrevin-GFP (visualized for GFP), andUas-Syt-HA (visualized with anti-HA antibodies). In addition, we generated a Uas-Syn-112 transgenic fly strain carrying a Uas-construct coding for the presynaptic protein Synapsin (visualized with anti-Synapsin antibodies; intrinsic Synapsin pattern is shown in Figs. 1 A′ and 6G,H). To restrict Synapsin toap-Gal4 neurons, we expressed Uas-Syn-1 in the absence of intrinsic Synapsin protein using theSyn97 mutant allele (ap-Gal4;Uas-Syn-112,Syn97 ). Animals homozygous for Syn97 are viable, and preliminary analyses at the light-microscopical, ultrastructural, and physiological levels have failed to reveal any obvious phenotypes (E. Buchner, unpublished results).
Targeted expression of presynaptic proteins suggests differential subcellular distribution of output synapses
When Uas-n-Syb-GFP, Uas-Syt-HA, orUas-Syn-1 is targeted to ap-Gal4 neurons, immunoreactivity can be detected in the synaptic neuropile (see below). However, staining is also localized to the nonsynaptic cortex, i.e., to Gal4-expressing cell bodies (Fig. 2, S) and in punctate patterns along cortical stretches of their primary neurites (Fig.2 C,E,F, open arrows). Thus, after targeted expression, synaptic proteins can be localized to extrasynaptic areas, and this is strongest inap:: Syt-HA animals (Fig.2 C,F). Similar results were obtained when mis-expressing Syt-HA pan-neuronally (elav:: Syt-HA) (Fig. 2 H′,asterisk). In late elav:: Syt-HAembryos, significant amounts of targeted Synaptotagmin were found in the cortex and along peripheral nerves (Fig. 2 H′,Cx and arrowheads). However, in the same animals, intrinsic Synapsin (Fig. 2 H) or intrinsic Cysteine string protein (data not shown) showed the usual restricted distribution to the neuropile and neuromuscular junctions (Fig.2 H, Np and asterisks). Thus, the general pattern of intrinsic output synapses seems unaffected by surplus amounts of mis-expressed synaptic proteins. Furthermore, the surplus mis-expressed protein seems to be either localized to transport vesicles (Ahmari et al., 2000) or deposited along extrasynaptic areas of neurites.
Within the neuropile, punctate patterns of targeted proteins reminiscent of output synapses can be seen along ap-Gal4processes. In most of these puncta, different targeted proteins seem to colocalize, as demonstrated via coexpression of Syt-HA and Syn-1 in the same Syn97 mutant animals (Fig.2 F′, bent arrows). However, detailed inspection of these animals reveals that Syt-HA, especially, can also occur alone (Fig. 2 F′′′, open arrowheads). Therefore, either not all output synapses are equally labeled by all tagged proteins or targeted proteins can localize randomly to ectopic, nonsynaptic sites. However, the overall pattern of targeted proteins shows clear tendencies of preferential localization, which can be obtained with different immunolabeling methods (staining with HRP/DAB/H2O2 revealed the same results; data not shown). First, strong and consistent accumulation of all targeted proteins is detectable in the median longitudinal fascicle (Fig. 2, L). Second, less consistent stain can be found along the transverse neurites (Fig. 2,T). This labeling of the transverse neurites is strong in ap:: Syt-HA animals (Fig.2 C,F) but weaker with Syn-1 or n-Syb-GFP. For Syn-1 it is even weaker at 18°C (where Gal4-activity is lower; data not shown). These variabilities suggest that the dotted stain in transverse projections might not represent true output synapses but surplus extrasynaptic protein, similar to stain found in cortical areas (see above and Discussion). Third, the only neurites that are consistently free of any of the targeted proteins (except for very occasional dots) are the side branches coming off the transverse primary neurite, suggesting that these side branches might be preferentially postsynaptic (Fig. 2, white arrows).
Taken together, our stainings suggest that output synapses seem to be localized preferentially to the median longitudinal fascicles formed by ap-Gal4 neurons (see Discussion). However, our findings just reveal tendencies that are obscured by potentially extrasynaptic localization of targeted proteins in ap-Gal4neurons.
Using transplantation-based mosaic analysis for the visualization of output synapses in the CNS ofDrosophila
Caveats regarding the Gal4/Uas approach are the limited number of suitable Gal4-lines and the artificial expression levels of targeted proteins resulting in restricted reliability for the identification of true output synapses. To overcome these problems, we established a new way of mosaic analysis that allows the visualization of intrinsically regulated presynaptic proteins in all neural cell lineages.
This technique is based on cell transplantation of single precursor cells (Prokop and Technau, 1993) (Fig.1 B,B′): a single cell from the neuroectoderm is transplanted from a labeled donor into an unlabeled host embryo at the early gastrula stage. There are 35 different neural precursors in each presumptive hemi-neuromere of the ventral nerve cord of Drosophila that can develop from these transplanted cells. All of their lineages have been shown to be reproducible, and detailed descriptions have been published (Bossing and Technau, 1994;Bossing et al., 1996; Schmidt et al., 1997, 1999). To visualize presynaptic sites, we use wild-type donor embryos carrying one copy of a Gal4 construct under the control of the ubiquitoustubulinP-enhancer and one copy of the Uas-CD8-GFPconstruct. Cells from these donors are transplanted into hosts that are homozygous for the Syn97 deletion (Fig.1 B). At the end of embryogenesis, their CNSs are removed, fixed, and stained with anti-CD8 and anti-Synapsin antibodies (Fig. 1 B′). Because only the transplantation-derived cells express the surface marker CD8, their projections can be visualized. Additionally, they are the only cells expressing the presynaptic protein Synapsin, which can be seen exclusively along CD8-labeled processes. Because Synapsin expression in these cells is based on the natural intrinsic control, its spot-like staining should represent true output synapses of these neurons.
Our analyses were performed at the end of embryogenesis, thus later than previous studies on identified embryonic neuronal lineages that were restricted to early and mid stage 17, i.e., stages in which cuticle is not yet fully developed (Schmidt et al., 1997, 1999). At the late embryonic stage, projections have grown further, and additional neurites have developed (Fig. 3). Furthermore, because of condensation of the nerve cord, cell bodies have often changed their position, the space between the neuropiles of both body halves is severely reduced, and the outlines of anterior and posterior commissures are less obvious. Nevertheless, a certain number of cell lineages could be identified with respect to published data for younger embryos (Table 1). Detailed descriptions and criteria used for their identification will be published elsewhere (but see examples in Fig. 3). Here we will focus on features relevant to the addressed topic.
Synapsin staining and Synaptotagmin staining of cell lineages are specific and reproducible
Strikingly, any cell lineages that were obtained more than once showed a reproducible distribution of Synapsin (Figs.4, 5). Minor variations were observed, and these might reflect either a degree of inherent variability of these cells or limitations of the method. Such variations never obscured the clearly conserved overall pattern of lineage-specific Synapsin distribution. We wondered whether Synapsin might reflect an incomplete pattern of synapse distribution. To test this possibility we made use of the presynaptic protein Synaptotagmin, which is essential for synapse function and for which antibodies and protein null mutations are available (Littleton et al., 1993;Südhof and Scheller, 2001). UsingSytAD4;Syn97 double-mutant embryos as hosts, we transplanted precursor cells fromtubP:: CD8 donors and obtained some cell lineages double labeled with anti-Synapsin and anti-Synaptotagmin. In these cell lineages, both proteins were generally colocalizing (Fig.4 I,J, white arrows andarrowheads). In some cases we could see rare spots labeled with Synaptotagmin alone; however, they intermingled with double-stained sites and therefore did not alter the principal pattern of potential output synapses (Fig. 4 J, white Vs). Some specimens were stained with only anti-Synaptotagmin and anti-CD8 antibodies. Resulting cases of identifiable lineages revealed a distribution of presynaptic spots that matched very well with images obtained from other examples of the same lineage labeled with anti-Synapsin and anti-CD8 (Fig. 4, Gvs H).
Thus, our transplantation-based mosaic analyses with two different presynaptic proteins reveal the same reproducible distribution of presynaptic sites within neuronal cell lineages ofDrosophila embryos and strongly suggest that output synapses can be reliably visualized with this technique.
Output synapses of cell lineages are restricted to specific neuropile areas
The distribution of output synapses within cell lineages can be restricted within the dorsoventral, ipsilateral–contralateral, anteroposterior, or mediolateral axis (Table 1). For example, lineages containing ventral unpaired median neurons (VUMs) (Fig.4 B,B′) contain two to six cells, which are located at the ventral midline of the nerve cord. Half of the VUM neurons are usually bifurcating motoneurons, and the other half are bifurcating interneurons (Bossing and Technau, 1994). The neurites of the VUM neurons can be seen from dorsal to median levels within the neuropile. Frontal presentations of such lineages demonstrate clearly that only projections in the more ventral area bear output synapses (Fig. 4 B′ii, white arrowheads), whereas dorsal neurites are free of Synapsin stain (Fig.4 B′ii, open arrowheads). A very similar result was observed for a further median cell lineage derived from the median neuroblast (MNB) (Bossing and Technau, 1994) (Figs.4 A, 5 C). In the NB5-2 lineage, anterior contralateral projections bear output synapses in the dorsal but not ventral plane of the neuropile (Fig. 4 G,H,open versus white arrowheads on left side in i vs ii, respectively). Similarly, the cell lineages derived from the precursors NB2-1, NB3-1, and NB5-3 show a dorsoventral-specific distribution of presynaptic sites within their neurites (data not shown).
Examples for ipsilateral–contralateral-specific distribution of output synapses are the cell lineages of NB1-1, NB3-2, NB5-3, and NB7-1. All lineages have neurites on both sides of the nerve cord. In the case of NB7-1, only contralateral projections bear output synapses (Fig.4 F,F′, white arrows andarrowheads). The same is true for NB3-2 and NB5-3, whereas NB1-1 restricts output synapses to ipsilateral projections (data not shown).
Anteroposterior-specific distribution of output synapses can be found in the lineages of NB6-1 and NB6-2 (Figs.4 D–E′, 5B). The NB6-1 lineage has neurites on the ipsilateral and contralateral side, which are directed anteriorly and posteriorly. Although on the contralateral side projections to posterior and anterior show Synapsin staining, anterior projections on the ipsilateral side are free of output synapses (Fig.4 D,D′, open arrowheads). An even more striking example of specific anterior–posterior distribution is the cell lineage of NB6-2. Its neurons have ipsilateral and contralateral neurites, and output synapses can be seen on both sides. Contralaterally, most projections turn posterior, and there is a sharp line behind which output synapses can be found (Fig.4 E,E′, asterisks) (less obvious in Fig. 5 B because of age of specimen as explained in legend). Except for a single focal accumulation of output synapses (Fig. 4 E,E′, white arrow), the anteriorly located contralateral neurites are devoid of output synapses.
The neural lineages of NB3-1 (data not shown), NB6-2, and NB2-1 show a mediolateral-specific distribution of output synapses. In NB6-2, an ipsilateral neurite projecting anterior along a intermediate longitudinal tract is free of Synapsin spots (Figs.4 E, 5 B, open arrowhead; less visible in Fig. 4 E′), whereas the more median ipsilateral anterior projections show output synapses. Vice versa, on the contralateral side, laterally located anterior projections show presynaptic staining, whereas more medially located projections do not (Figs. 4 E,E′, Fig. 5 B,open vs white arrow on the left). Neurons of the NB2-1 lineage project to the contralateral side where they form medial and lateral arborizations. Only the laterally located projections show presynaptic spots (Fig. 4 C,C′,white arrow vs open arrowhead).
Output synapses are reproducibly restricted to compartments of neurites
Taken together, our analyses demonstrate that some neuronal processes harbor output synapses, whereas others do not. Furthermore, our lineages demonstrate that those neurites that have output synapses have them restricted mostly to reproducible neuritic compartments. For example, in the cell lineages of neuroblasts 2-1, 3-1, 3-2, 5-2, 5-3, 6-1, 6-2, and 7-1, neurites enter the neuropile ipsilaterally and cross the midline. Contralaterally the neurites reach into defined areas of the neuropile and form branches. Often output synapses can be seen only on these contralateral branches, whereas the transverse primary neurites are void of stain (Figs. 4, 5, open straight andbent arrows). This phenomenon is most striking for the lineages of neuroblasts 2-1 and 6-2, because in these lineages more or less isolated neurites can be seen that bear accumulations of output synapses only at their very tips (Figs.4 C,C′,E,E′, 5B, white arrows). Other examples become very obvious when we scroll through the different focal planes of our preparations, but they are mostly obscured when pictures are compressed into one layer (data not shown). In those examples mentioned for NB2-1 and 6-2, presynaptic arbors are restricted to a small area (Figs.4 C,C′,E,E′, 5B, white arrows), indicating that these cells represent local interneurons. In many other cases the presynaptic arbors coming off transverse neurites seem to lie predominantly in longitudinal tracts and often extend across neuromere borders (potential intersegmental interneurons) (Figs. 4, 5, white arrowheads).
Our lineage analyses show clearly that output synapses are often restricted to reproducible compartments on individual neurites. Preferentially primary transverse neurites are void of output synapses. Our transplantation experiments are in agreement with our observations from the experiments with ap-Gal4 and encourage use of both methods for future investigations of mechanisms underlying synapse formation and localization in the Drosophila CNS.
Mechanisms underlying formation of output synapses in the CNS and at neuromuscular junctions might differ
Having learned about basic principles of synapse distribution in central neurons, we have begun to apply this knowledge and our transplantation strategy for the genetic analysis of synaptogenesis in the CNS. Thus, we ask whether mutations affecting the development of NMJs (Fig. 1 C) might also interfere with synaptogenesis in the CNS. Mutations of the cytoskeletal interacting factor Kakapo (also called “Short stop”) show a severe reduction of output synapses at embryonic motor terminals of Drosophila (Prokop et al., 1998) (Fig. 6, J vsH). Furthermore, kakapo mutations show defects in the CNS (reduced neurites and failure to compartmentalize the transmembrane protein Fasciclin 2) (Prokop et al., 1998), but these CNS phenotypes have not been studied at the synaptic level. To perform such analyses, we generated recombinant fly stocks that allow labeling and identification of kakapo-mutant cell lineages (tubP-Gal4,kakDf(2R)MK1/Uas-CD8-GFP,kaksf20 ) (Fig. 5 G). Using these embryos as donors for transplantations andkak+;Syn97 embryos as recipients, we obtained cell lineages of NB5-2 (Fig.5 D), NB6-2 (Fig. 5 E), MNB (Fig.5 F), NB1-1 (data not shown), and NB3-2 (data not shown). For reasons yet unknown, most examples of mutant lineages (except NB6-2) displayed an extremely weak and discontinuous CD8 staining so that the precise shape and extent of neurites could not be analyzed in detail. However, all of these lineages showed strong Synapsin staining, making it easier to interpret the location of the presynaptic projections. Output synapses seemed clearly present as revealed by strong dots of Synapsin staining, although they appeared slightly blurred and extended compared with wild type, and occasionally Synapsin seemed to aggregate into larger spots than usual. However, compartmentalization of output synapses seemed normal with respect to the criteria described before for the respective wild-type lineages. More surprisingly, the amount of output synapses was not different from wild-type lineages when single Synapsin spots were counted in our specimens nor was the area over which the synaptic spots were distributed (Table 1) (for one special exception, see Fig.5 E, X). In contrast, motoneuronal terminals and the number of neuromuscular output synapses in these donor embryos were strongly reduced, in agreement with previous descriptions (Fig. 6, J vs H, arrows) (Prokop et al., 1998). One explanation for the mild phenotypes in the CNS might be that the analysis of single mutant cell lineages might be rescued via some non-cell-autonomous contribution of the surrounding wild-typic host tissue. However, our analyses ofkakaposf20/kakapoDf(2R)MK1 mutant CNS likewise showed a rather normal amount of Synapsin staining and even maintained segmentally repeated pattern elements described for the wild-type neuropile earlier in the text (Fig. 6, I vsG). Two further kakapo-mutant alleles (short stop3 ) (Fig.6 M,N) andkakapoHG25 (data not shown) similarly showed affected NMJs but normal amounts of synaptic staining in the neuropile, regardless of the presynaptic marker that we used for our analyses.
Hence, kakapo-mutant embryos show strong reduction of output synapses at the NMJ but surprisingly little effects on the distribution and amount of presynaptic stain in the CNS. This suggests that molecular mechanisms of synaptogenesis might differ to a certain degree in the CNS and at motor terminals, a possibility that needs to be considered for future work. Furthermore, our experiments withkakapo-mutant embryos show that the transplantation technique can be combined with genetic approaches and represents an important method for future analyses of mechanisms underlying synapse formation in the CNS.
Output synapses are compartmentalized on neurites of embryonicDrosophila neurons
Here we show for the first time the restricted localization of output synapses to specific neuritic compartments in the CNS ofDrosophila and that they are distributed in reproducible patterns. Our results suggest the existence of at least three different types of neurites or neuritic compartments. First, a fraction of neurites clearly harbors output synapses that have the tendency either to be arranged into longitudinal tracts (Figs. 4, 5, white arrowheads) or to form local arborizations (Figs. 4, 5,white arrows).
Second, especially transversely oriented (primary) neurites leading toward output areas are usually void of Synapsin in our transplantation-derived specimens and might represent transport highways. This might explain the random dots of staining when synaptic proteins are targeted to apterous-Gal4 (Fig. 2,T), which might be attributable to the fact that excess amounts of targeted protein are traveling in these transverse neurites (Ahmari et al., 2000) or get deposited on their way. We are not aware of data showing that vertebrate neurons might compartmentalize their output synapses along their axons and nondendritic neurites (Peters et al., 1991), although one would perhaps expect that they do. For example, parallel fibers in the cerebellum might first project through the Purkinje cell layer but not form output synapses before they reach the molecular layer. On the other hand, cultured hippocampal neurons show no obvious compartmentalization of output synapses in nondendritic processes (Rao et al., 2000), but this might not reflect the situation within the intact brain tissue.
Third, our apterous-Gal4 experiments revealed that branches coming off the primary neurite on its way through the neuropile were basically void of the mis-expressed presynaptic proteins. Also, unstained neurites could be found in the transplantation-derived specimens (Figs. 4, 5, open arrowheads). Such neurites might represent true postsynaptic structures, as found similarly for neurons in larger insects (Killmann et al., 1999) (for review, see Burrows, 1996). Indeed, our present investigations show that certain neurites ofDrosophila neurons share several features with postsynaptic vertebrate dendrites (unpublished observations).
So far, statements about compartmentalization of neurites could be made only for repetitively organized neural tissues of the CNS. For example, the distribution of synaptic markers in the repetitively organized optic lobes of flies reveals that primary neurites have stretches of nonsynaptic surfaces because they project through synapse-free areas among lamina, medulla, and lobula/lobula plate (Hiesinger et al., 1999;Meinertzhagen and Sorra, 2001). Also, output synapses on the efferent projections of motoneurons are clearly restricted to the neuromuscular terminal (Fig. 6 H). The situation was unclear for more diffuse regions like the neuropile of the ventral nerve cord, and our studies are the first to uncover the synaptic distribution at the level of individual projections.
Two different strategies to visualize synapses within individual neurites
To gain insight into the subcellular organization of neurons in the Drosophila CNS, we used light-microscopic analyses of genetic mosaics, either obtained from transplantations or with the Gal4/Uas-system. For our lineage analyses we modified existing techniques (Prokop and Technau, 1993) with respect to the genotypes of host and donor embryos and the stage of analysis (Fig. 1). This allows the visualization of the true intrinsic expression and localization profile of presynaptic proteins such as Synapsin and Synaptotagmin. This cannot be achieved with genetic methods of clonal analysis that require heterozygous backgrounds (for example, the mosaic analysis with a repressible cell marker technique) (Lee and Luo, 1999). In contrast, our approach requires a homozygous mutant condition in which the antigen is absent in all but the implanted cells. Our transplantation-based strategy can be used in different ways. First, as demonstrated for kakapo, either donor or host embryos can carry mutant alleles of potential synaptogenic genes, allowing tests for cell-autonomous versus signal-dependent requirements of these genes. Second, donor or host embryos can mis-express genes of interest to test their function in greater cellular detail. Third, the stage of analysis is not restricted to the embryo (Prokop and Technau, 1991). Fourth, our method has the potential to be used for the study of the subcellular localization or cell specificity of any other neural gene, provided that good antibodies are available, and mutant alleles, which eliminate the antigen in host embryos (but do not cause severe phenotypes). A disadvantage of the technique lies in the randomized choice of the precursors, which are transplanted. The choice can be narrowed down only within the dorsoventral axis of the neuroectoderm (Prokop and Technau, 1993), thus restricting the number of different lineages that will be obtained (e.g., most of our experiments were restricted to the ventromedian half of the neuroectoderm). However, once lineages are identified and described, they can be used as templates for further studies, as we have successfully demonstrated for the analysis of kakapo-mutant cell lineages.
The Gal4/Uas-based strategy (Brand and Perrimon, 1993) of synapse visualization uses targeted expression of presynaptic proteins and their detection via epitope tags (n-Syb-GFP, Syt-HA) or via genetic removal of the respective intrinsic protein from the whole nervous system (Uas-Syn-112 inSyn97 background) (Fig. 2). The strength of this technique is the ease of its application and the amount of data that can be produced in a short period of time (Ito et al., 1998;Blagburn et al., 1999; Ahmari et al., 2000; Vosshall et al., 2000). The disadvantage of mosaic analyses via targeted expression of synaptic proteins lies in the low number of available Gal4 lines with restricted neuronal expression patterns in late embryos or larvae. This limits the experimental possibilities, and results can hardly be generalized. Furthermore, experimenters will have to take into consideration the fact that extrasynaptic staining might occur, especially when constitutively active Gal4 drivers cause overproduction of the protein (Fig. 2). In contrast, intrinsically regulated synaptic proteins are restricted to the neuropile and NMJs in mature embryos (Figs.1 A,A′, 6G,H). Nevertheless, our analyses in the ventral nerve cord of apterous-Gal4 embryos and larvae suggest that preferential output areas of Gal4-expressing neurons can be visualized. In agreement with the model derived from our transplantation studies, output areas of apterous-Gal4neurons seem to accumulate within the longitudinal median fascicle, whereas the variable and construct-dependent localization along the transverse primary neurite is more likely to represent extrasynaptic surplus protein. Hence, Gal4-based visualization of synapses has the potential to be used in combination with genetic or experimental manipulations or for mutational screens.
What might be the mechanisms of compartmentalization?
On the one hand, mechanisms involved in the compartmentalization of output synapses along neurites must depend on intracellular components organizing the cortical cytoskeleton. Indeed, inDrosophila motoneurons, presynaptic structures can assemble in the absence of extracellular contacts (i.e., without postsynaptic muscles), demonstrating the existence of cell-autonomous mechanisms (Prokop et al., 1996). At NMJs, Kakapo seems to represent such an intrinsic factor because there are strong indications that Kakapo functions cell autonomously (Prokop et al., 1996; Lee and Luo, 1999). However, kakapo mutant alleles affecting NMJ differentiation failed to demonstrate a similar requirement of Kakapo for the number and localization of CNS output synapses (Fig. 6). This indicates potential differences of synaptogenetic mechanisms in the CNS and at NMJs. Such an scenario would not be unlikely, given the fact that most central synapses are cholinergic or GABAergic, whereas NMJs are glutamatergic (Johansen et al., 1989; Takagawa and Salvaterra, 1996; our unpublished results). Also in vertebrates, developmental mechanisms seem to differ at central and neuromuscular synapses (Serpinskaya et al., 1999).
Intracellular components required for the distribution of output synapses might be (partially) orchestrated by extracellular cues (Suter and Forscher, 2001). Thus, in the absence of such extrinsic cues, synapses might be formed but mislocalized. A paradigm for response of intrinsic synapse assembly mechanisms to extracellular cues was established in hippocampal cell cultures where axonal transport vesicles carrying various presynaptic proteins get trapped at points of contact with dendrites, thus seeding new output synapses (Ahmari et al., 2000). In contrast to randomly provided signals in cell culture, extrinsic signals in the stereotypically structured CNS could be provided in an ordered manner. This could explain the reproducible localization of output synapses that we observed. Cell cultures for Drosophila in which extrinsic versus intrinsic mechanisms can be tested are established in our laboratory and presently analyzed in this context (unpublished observations).
One attractive thought is that mechanisms patterning the CNS during earlier development might have an impact on the distribution of output synapses. For example, cell-surface molecules involved in pathfinding aspects in grasshoppers or Drosophila seem to be compartmentalized during earlier development (Goodman and Doe, 1993;Rajagopalan et al., 2000), and mechanisms underlying compartmentalization of output synapses might be related to these mechanisms. No mechanisms of compartmentalization have been identified so far, but Drosophila provides efficient tools and techniques to address them.
This work was supported by grants from the Deutsche Forschungsgemeinschaft (DFG) (PR605/1 and PR605/2) and the German Israeli Foundation (I 073-203.05/98) to A.P. and by a DFG grant (Bu566/6) to E.B. We express our gratitude to Gerhard M. Technau for advice, helpful discussions, criticism on this manuscript, and provision of lab space and equipment. We are grateful to our colleagues Hugo Bellen, Liqun Luo, Mani Ramaswami, Iain Robinson, John Thomas, and Konrad Zinsmaier and the Bloomington Stock Center for providing fly stocks and antibodies, as named in detail in the text. Furthermore, we thank Matthias Landgraf and Natalia Sánchez-Soriano for constructive and helpful criticism on this manuscript.
Correspondence should be addressed to Andreas Prokop, Institute of Genetics, University of Mainz, J.-J.-Becherweg 32, D-55128 Mainz, Germany. E-mail:.