The Drosophila tumor suppressor gene, dlg, is involved in structural plasticity at a glutamatergic synapse

Dynamic dlg expression at type I boutons

Our previous studies demonstrated that DLG is expressed at both the pre- and postsynaptic membrane during the last stage of larval development [19]. Moreover, in the presence of abnormal dlg, postsynaptic structure was disrupted. To understand the function of dlg at glutamatergic synapses, we first examined the developmental pattern of DLG expression during synaptogenesis and synapse maturation (Figs 2,3). During embryonic stages 15, 16, and early-to-mid stage 17, DLG was expressed lightly in most neurons (data not shown). During this period, many muscles became innervated (or contacted by motorneurons), but little or no label was found at boutons or postsynaptic muscles (Fig. 2a,b). In contrast, during late stage 17, about 2 hours before hatching, DLG immunoreactivity was localized primarily in the axons and at type I boutons (Fig. 2c). Double staining of these late stage 17 preparations with the nerve-specific antibody, anti-horse radish peroxidase (HRP) [31], and with anti-DLG antibodies revealed that DLG was present inside type I boutons (Fig. 2c, arrowhead). However, the developing bouton border and their filopodia were devoid of DLG immunoreactivity (Fig. 2c, arrow).

First instar larvae were studied at two periods — within 20 minutes of hatching (early first instar), and at about 8–12 hours after hatching (mid first instar). In contrast to late embryo, DLG immunoreactivity was concentrated at the border of synaptic boutons in early first instar larvae (Figs 2d and 3a). Double-labeling experiments with anti-HRP and with anti-DLG antibodies suggested that staining at the bouton borders represented DLG expression at both pre- and postsynaptic sites (see below). At this stage, anti-HRP antibodies are known to stain axons and presynaptic boutons [32]. However, we found that DLG immunoreactivity co-localized only partially with HRP immunoreactivity at the edge of the bouton (yellow label in Fig. 2d), and that it extended into the muscle junctional area, which was devoid of HRP immunoreactivity (green label and arrow in Fig. 3d). At this stage, the DLG label appeared as a crescent-shaped area, at the border of each bouton, and was undetectable in the axon (Fig. 5a).

During later larval stages, DLG immunoreactivity became progressively localized around the entire periphery of the bouton (Figs 2e and 3b–d). As with the first instar, double-labeling experiments indicated that some of the anti-DLG label colocalized with anti-HRP-immunoreactivity at the bouton border. However, DLG immunoreactivity became increasingly extensive in the postsynaptic junctional muscle region. This observation is consistent with our previous immunoelectron microscopic study showing that anti-DLG immunoreactivity is concentrated in the SSR, and less prominently at the presynaptic membrane [19].

The results described above indicate that DLG expression is dynamic, being concentrated first in axons and presynaptic boutons, and later at presynaptic boutons and postsynaptic junctional sites. These findings suggest that DLG may have a function first in the motorneuron, well after synaptogenesis, and later in the muscle for the proper formation of the SSR. We therefore examined the late-embryonic and post-embryonic development of type I synapses in wild-type and in dlg mutants.

The SSR is formed during the first larval instar, and increases in complexity during late larval stages

The most prominent phenotype of dlg type I synapses was an abnormal SSR appearance during the last larval stage [19]. We therefore looked at the development of the SSR at the electron microscopic (EM) level in wild-type and in dlg mutants, to determine whether structural abnormalities in the mutant were the result of disrupted development or disrupted maintenance of type I synapses.

Transmission EM examination of semi-serial thin sections through type I boutons at larval stages demonstrated that the SSR began to develop during the first larval instar, from the junctional muscle membrane. At this stage, the postsynaptic membrane was very simple and unfolded, forming a depression or gutter over which lay the bouton (Fig. 4a). The presynaptic bouton at this stage was smaller than third-instar type I boutons. However, like the third-instar bouton, it was circular to oval in cross-section, and was filled with 40 nm clear synaptic vesicles. Moreover, the presumed active zones, which appeared as dense ‘T’ shaped structures associated with a cluster of synaptic vesicles, were already visible (small arrow in Fig. 4a).

During mid-first instar stage (8–12 hours after hatching), the junctional muscle membrane began to invaginate at several regions, forming long finger-like structures — the initial folds of the SSR — which partially surrounded type I boutons (data not shown).

During the second instar stage, the SSR became more extensive, having an increased number (about 3–4) of membrane layers around the bouton (Fig. 4b). In contrast to later stages, however, these membrane layers were unfolded and concentric to the presynaptic membrane. In addition, many boutons were still localized on the surface of the muscle and were not completely surrounded by the SSR. Complete surrounding of boutons by the SSR became apparent during the late second instar stage.

By the early third instar stage, the SSR began to acquire its final, wandering third instar semblance. Many layers of membranes now surrounded the boutons, and these layers began to fold, losing their parallel arrangement with the presynaptic membrane, and acquiring the elaborate appearance of third instar type Ib SSR (Fig. 4c). Figure 5d shows a diagrammatic representation of these developmental changes at the SSR.

To quantify the increase in membrane surface during the larval stages, the SSR cross-section at the bouton midline was traced for each bouton, scanned into a computer, and the SSR membrane length and the presynaptic bouton cross-sectional area measured (Fig. 5a,c). We found that the SSR cross-sectional length increased in an exponential fashion, following equation 1:

where L_SSR is the mean SSR length (μ), t is the time after hatching (hours), and R is the correlation coefficient for equation 1 (Fig. 5a). Comparisons of the SSR length with the muscle volume indicated that both followed an exponential time course (Fig. 5b). The change of muscle volume through the larval stages could be described by equation 2 (Fig. 5b):

where Vm is the volume of the muscle (μm³). The time constant of Vm, however, was 60 % larger than the SSR length–time constant. These results indicate that the increase in SSR length may be related to the dramatic increase in muscle size (154-fold in volume from first to third instar) during larval development. It has been shown previously that the number and size of boutons also increase exponentially during this stage [4]. Although a cause–effect mechanism cannot be established at this time, these results show that neuromuscular junctions expand continuously during the larval period as the target muscles grow. This neuromuscular junction expansion is the result of the growth of presynaptic endings (increase in number and size of boutons [4]) and, as we show here, of an increase in the postsynaptic surface.

The SSR does not expand normally in dlg mutants

To determine when the dlg mutant phenotype was first apparent during development, we performed similar ultra-structural and morphometric analyses of dlg^v59/Df mutants (Figs 4,5). In these dlg/Df mutants, the SSR expanded more slowly than in wild-type (L_SSR(t) = 18.8 × e^0.02t; R = 0.98), never reaching the final length of wild-type SSR.

For example, during the third instar, SSR membranes in dlg/Df mutants had an unfolded appearance that resembled wild-type second instar SSR (Fig. 4d–f). Similarly, the SSR cross-sectional length was significantly smaller (p < 0.001) during late second to third instar larval stages (Fig. 5b,c).This difference could not be accounted for by a difference in bouton area or muscle volume, which were similar to wild-type (Fig. 5b,c). These observations are summarized in a diagrammatic form in Figure 5d. These results also demonstrate that the signaling mechanisms that regulate the size of the muscle junctional membrane during target muscle growth are altered in the mutants.

DLG is required in embryos and in mid-to-late larvae for development of normal synaptic structure

The expression of DLG in the presynaptic bouton during late embryogenesis and in both the pre- and postsynaptic cell during larval stages suggested that dlg may function throughout synapse maturation. To determine when dlg is required for proper development of type I boutons, we examined a temperature-sensitive allele of dlg [23]. At the permissive temperature (18 °C), dlg ^ts/Df behaved as an hypomorph; for example, dlg ^ts/Df mutants still developed imaginal disc tumors, but these tumors were very small. At this permissive temperature, type I boutons were similar to those in wild-type at both the light and electron microscopical level (Fig. 6a,c,d,f). In wild-type, the mean cross-sectional SSR length was 370.9 ± 22 μm, as compared with 364.9 ± 19 μm in dlg ^ts/Df at permissive temperature.

At the restrictive temperature (29 °C), dlg ^ts/Df was similar to the genetically null allele, dlg^m52/Df, or to dlg^v59/Df. Tumors in the imaginal discs and brain were very large, and the SSR at type I boutons was poorly developed (Fig. 6b,e). In dlg ^ts/Df reared at restrictive temperature, SSR length was 169.2 ± 19 μm, which was similar to dlg^v59/Df (172 ± 13 μm).

The reduction in SSR length and complexity, determined by the ultrastructural studies, was paralleled by changes at the light microscopical level that could be easily observed using anti-DLG immunocytochemistry. Previous immunoelectron microscopical studies revealed that anti-DLG antibodies strongly stained the SSR and, to a lesser extent, the presynaptic membrane [19]. In dlg ^ts/Df at 18 °C, as in wild-type, DLG immunoreactivity, observed at the light- microscopical level, appeared as a thick doughnut-shaped structure that surrounded type I boutons (Fig. 6d,f). This structure was present, but clearly thinner in dlg ^ts/Df reared at 29 °C, as it was in dlg^v59/Df (Fig. 6e; [19]). These results show that the structural changes in dlg ^ts/Df mutants can be detected at the light microscopical level using anti-DLG antibodies.

We then shifted dlg ^ts/Df animals and wild-type controls to 29 °C at different stages of embryonic and larval development to determine whether there were critical stages when DLG was required for normal development of the SSR. The morphology of type I boutons, visualized by anti-DLG immunocytochemistry, was then examined at the wandering third instar stage, and blindly scored by four observers according to the criteria described in the Materials and methods section.

Figure 7 shows the percentage of samples with a mutant phenotype at type I synapses in animals subjected to different temperature-shift paradigms. There were two distinct stages where low DLG levels affected type I bouton structure in the third instar stage. The most dramatic effect of dlg suppression was during embryonic stages —most dlg ^ts/Df larvae pulsed during the embryonic stage died before reaching wandering third instar stage, and about 85 % of the surviving animals of this group displayed a mutant phenotype at third-instar type I boutons (Fig. 7). This effect was not due to temperature changes, because about 93 % of wild-type controls had a wild-type phenotype at type I synapses. Interestingly, a 29 °C temperature shift during the first instar, immediately after embryogenesis, was without effect in dlg ^ts/Df larvae. We conclude from these experiments that dlg is required during embryogenesis, possibly after the period when synapses are first formed.

A second, though less dramatic, requirement of dlg was found during larval stages: the later during larval development that the temperature was shifted to 29 °C, the higher the percentage of samples with a mutant phenotype. Maximum effect was obtained during mid second to third larval instar, resulting in about 60 % of samples with a mutant phenotype.

The findings that DLG was required during embryonic stages for normal development of the third instar type Ib bouton suggested that dlg may be important at earlier stages of muscle innervation. To investigate whether abnormal DLG would result in an abnormal differentiation of synaptic boutons at embryonic stages, we examined the development of innervation at muscles 6, 7, 12 and 13, in wild-type, dlg^v59/Df and dlg^m52/Df from stages 15–17 (Fig. 8). For these studies, we used two markers of growing axons and synaptic boutons — anti-Fasciclin II and anti-HRP. No differences were found either in pathway that axons follow, or in growth-cone morphology, or in the timing at which growth cones reach their target and start to differentiate into synaptic boutons. These results were not surprising, as studies of DLG expression had shown that dlg is not expressed at the presynaptic boutons until late embryogenesis, a few hours before the beginning of the larval period.

Fly stocks

Flies were reared using standard procedures at 25 °C unless otherwise indicated. The following stocks were used for these studies: dlg^HF321 (referred to as dlg^ts), a temperature-sensitive allele of dlg [23,39]; dlg^v59, a mutant allele that deletes a third of the guanylate kinase domain [15]; and Df(1) N71 (referred to as Df), a deficiency that uncovers the dlg locus. Mutant dlg stocks were maintained over the balancer FM7 and a Y chromosome containing a wild-type copy of dlg. The deficiency stock was maintained using a duplication, Dp(1:2)v[65b] (referred to as Dp), containing a wild-type copy of dlg in the second chromosome. All experiments were performed on dlg/Df mutants, obtained by mating dlg/FM7 females to Df;Dp/+ males. The female dlg/Df progeny can be identified by the presence of tumors in the imaginal discs and CNS. As wild type, the strain Canton-S (CS) was used.

Developmental studies and immunocytochemistry

Larvae were staged using the morphology of anterior spiracles and mouth hooks, and their age and length. Early first instars were dissected within 20 min after hatching. Studies in embryonic stages, before cuticle formation, were done using the fillet preparation [32]. To distinguish mutant samples from heterozygotes, embryos were double-stained with anti-DLG antibodies and either anti-Fasciclin II monoclonal antibodies (kindly provided by C.S. Goodman), or anti-HRP. Mutant embryos could be clearly identified by the lack of DLG-immunoreactivity in epithelial cell borders (cuticle-secreting cells and salivary gland cells) which stain brightly in wild-type. Studies in late embryos, after cuticle formation, were performed by manual dechorionation and devitelinization of embryos, followed by dissection of body walls as in larval stages [4]. Immunocytochemistry and confocal microscopy were performed as in [19]. For developmental measurements of muscle volume, body wall muscle preparations from each stage were stained with FITC-conjugated phalloidin, and volume of muscle 7 (Vm) was estimated assuming an elliptic muscle cross-section, and by measuring muscle thickness (a), muscle length (L), and muscle width (b) at the confocal microscope, according to the equation Vm = L2π√(a² + b²)/2.

Electron microscopy

Samples were prepared for TEM as in [25]. For studies in first and second larval instars, samples were fixed for 30 min to 2 h in fresh Trump’s fixative. Longer fixation times before osmium treatment at these stages resulted in poor preservation of membranes. All studies were done on the ventral longitudinal muscles 6, 7, 12 and 13 of the second or third abdominal segments using semi-serial sections. To distinguish dlg^v59/Df from dlg^v59/FM7 females before the third instar larval stage (when visualization of imaginal disc tumors is difficult), the intestine from each sample was dissected out, and separately stained for anti-DLG immunocytochemistry. While intestine epithelia from dlg^v59/FM7 displayed immunoreactivity at cellular borders, intestine epithelia from dlg^v59/Df were devoid of immunoreactivity, allowing precise identification of dlg^v59/Df samples.

For morphometric analysis of boutons, EM negatives were printed at 30 000 × final magnification, and the subsynaptic reticulum (SSR) traced. SSR drawings were scanned into a computer, and analyzed as in [19] using the program NIH Image 1.57. The number of wild-type samples and type Ib boutons used for the EM morphometric analysis was: first instar, 2 samples (16 boutons); second instar, 6 samples (26 boutons); early third instar, 2 samples (13 boutons); wandering third instar, 3 samples (10 boutons). The number of dlg^v59/Df mutant samples was: first instar, 2 samples (16 boutons); second instar, 4 samples (28 boutons); early third instar, 2 samples (13 boutons); wandering third instar, 3 samples (10 boutons). Results were expressed as mean ± S.E.M.

Temperature pulses

Broods (from the parental cross dlg^ts/FM7 × Df/Y; Dp/+) were maintained in an 18 °C incubator, and shifted to a 29 °C incubator at different stages of development. Identical temperature pulses were performed in wild-type. As temperature pulse shifts affect the length of each larval stage, and because there is some variability in absolute developmental times, we used the molts as markers for transition between instars. To determine molting stages, plates containing the larvae were examined for the presence of shed mouth-hook parts and larvae in active molting behavior. In general, the duration of development (from egg laying to wandering third instar) at 29 °C was 4 ± 2 days for dlg^ts/Df, and 4 ± 1 days in controls. At 18 °C it was 15 ± 3 days in dlg^ts/Df, and 10 ± 2 days in controls. The duration of development in all the temperature paradigms was 13 ± 1 days for dlg^ts/Df, and 10 ± 1 days for controls. For temperature pulses involving embryonic stages, parents were placed at 29 °C for 5 h prior to transferring to a new food vial for examination of the progeny. dlg^ts/Df females were identified by dissecting the larvae and looking for the presence of tumors in the imaginal discs. We found that, even at 18 °C, dlg^ts/Df had small but visible tumors in the imaginal discs. For analysis of boutons, images collected by confocal microscopy were blindly and independently scored by 4 naive observers according to the following criteria. Observers were first presented with 5 different representative examples of wild-type and dlg/Df type Ib bouton strings. Then they were instructed to score the experimental and control type Ib in a scale of 1–3 with regard to two characteristics of type Ib boutons which are affected in the mutants: the thickness of the bouton rim stained by anti-DLG antibodies; and the presence of perisynaptic network. Samples with a score of 3 or less were considered as ‘wild type’ and samples with a score of more than 3 as ‘mutant’. The percentage of mutant samples in each experimental and control group was expressed as the average percentage found by each observer ± S.E.M.