As previously demonstrated 16171819, DLG is abundantly expressed throughout the postsynaptic membrane surrounding presynaptic motor axon terminals (aka 'boutons') (Fig. 1A–B). DLG appears distributed throughout the postsynaptic membrane; there are no discernible DLG-positive domains smaller than the size of a bouton. In contrast, immunoreactivity for NMJ glutamate receptors has been shown to be restricted to specific postsynaptic domains directly opposite presynaptic active zones 2635. This restricted immunoreactivity is visible as distinct clusters within the bouton-wide area delimited by DLG staining (Fig. 1C–D). Thus, not all postsynaptic DLG appears associated with glutamate receptors. We cannot determine by light microscopy whether all glutamate receptors colocalize with DLG, although our staining is consistent with that conclusion.

In Drosophila embryos and larvae, the intersegmental nerve branch b (ISNb) innervates the ventral longitudinal muscles of each abdominal hemisegment 36. In the confocal images shown in Fig. 2, ISNb is visualized using anti-HRP antibodies, which stain all neuronal membranes (green). Three NMJs on four muscles (not stained) are shown in each image (Fig. 2A–B). Ventral longitudinal muscles 7 and 6 are innervated via a NMJ that lies in the cleft between the two adjacent muscles. Muscles 13 and 12 are innervated by arborizations distal to the 7/6 NMJ. Each of these body wall NMJs contains multiple clusters of postsynaptic glutamate receptors that can be visualized using antibodies specific to either GluRIIA (Fig. 2A) or GluRIIB (Fig. 2B).

To determine whether DLG is required for clustering of postsynaptic glutamate receptors, we visualized NMJ glutamate receptors in control and DLG mutant embryonic NMJs using GluRIIA and GluRIIB specific antibodies 30. To manipulate DLG levels, we used embryos homozygous for the mutation dlg^X1–2. In dlg^X1–2 mutants, the S97N isoform of DLG, which is predominantly expressed in neurons and muscle, is reduced to undetectable levels 37. Other isoforms of dlg (Drosophila expresses at least five) are expressed only at extremely low levels (approximately 5% normal) 37. In addition, all isoforms (including S97) are truncated such that the C-terminus and GUK domains are completely removed 20.

Both control and dlg^X1–2 mutant NMJs contain highly visible clusters of GluRIIA-containing receptors (Fig 2A; magenta). In Drosophila embryonic NMJs, the cluster area is directly proportional to the number of functional postsynaptic receptors measured using patch-clamp electrophysiology 25. GluRIIA cluster area does not differ significantly between control and dlg^X1–2 mutants (control cluster area = 0.68 ± 0.03 μm², N = 103 clusters from 10 embryos; dlg = 0.75 ± 0.03 μm², N = 99 clusters from 16 embryos; P = 0.15). Immunoreactivity for GluRIIB, on the other hand, appears dramatically decreased in dlg^X1–2 mutants compared to controls (Fig. 2B). Indeed, GluRIIB cluster size is significantly decreased in dlg^X1–2 mutants compared to controls (control cluster area = 0.45 ± 0.03 μm², N = 88 clusters from 6 embryos; dlg = 0.31 ± 0.02 μm², N = 57 clusters from 6 embryos; P < 0.001). Cumulative frequency histograms of GluRIIA and GluRIIB cluster sizes (Fig. 2C) represent the entire distribution of cluster sizes measured in control and mutant embryos. Fig. 2C shows that the distribution of GluRIIA cluster sizes is almost identical in control and dlg mutants. The GluRIIB cluster size curve in dlg mutants, however, is shifted toward smaller values. The largest shift occurs in the section of the curve where cluster sizes are largest, suggesting that the largest GluRIIB clusters are preferentially lost in dlg mutants. However, the smallest clusters approach the resolution limit of light microscopy, where reductions in object size are no longer detectable. Thus, the reduction in small cluster size is probably underestimated, and the average decrease of GluRIIB in dlg mutants may be larger than is measurable by immunocytochemistry.

The immunocytochemistry in Fig. 2 suggests that dlg^X1–2 mutants specifically lose receptors that contain GluRIIB, but do not lose receptors containing GluRIIA. GluRIIA null mutants are viable, but mEJP amplitudes are smaller, glutamate receptor channel open times are reduced, and receptors show decreased sensitivity to the GluR antagonist argiotoxin 636 31. GluRIIB null mutants are also viable, but show no significant change in receptor function, suggesting that either the GluRIIB subunit plays a lesser role in channel function, or that the majority of native receptors lack GluRIIB.

To confirm our immunocytochemical results, and explore the functional changes resulting from loss of GluRIIB-containing receptors, we used electrophysiology. First, we compared single glutamate receptor channel properties in control and dlg mutant embryonic muscle 6 (Fig. 3A–B). Because Drosophila glutamate receptor conductance is relatively large and embryonic muscle input resistance is relatively high, single glutamate receptor channel currents are visible during the falling phase of some spontaneous synaptic events (Fig. 3B). DiAntonio et al. 31 showed that extrasynaptic larval muscle glutamate receptors in GluRIIB null mutants have slightly larger single channel currents (8.8 pA and 9.2 pA at -60 mV, for wild-type larvae and GluRIIB null mutants, respectively). We saw a similar, but larger increase in synaptic receptor single channel amplitudes in dlg mutant embryos (Fig. 3A; control = 9.3 ± 0.7 pA at -60 mV, N = 17; dlg = 14.1 ± 0.06 pA at -60 mV, N = 42; P < 0.001).

DiAntonio et al. 31 also examined single channel kinetics in the absence of GluRIIA or GluRIIB. Although loss of GluRIIA resulted in a dramatic decrease in average open channel times, loss of GluRIIB did not result in any significant change in open channel duration, compared to wildtype. If dlg mutants selectively lose GluRIIB, but not GluRIIA, then there should correspondingly be no change in single glutamate receptor channel kinetics in dlg mutants. Consistent with this, we observed no change in average duration of single channel currents visible during the falling phase of spontaneous synaptic currents (control channel open time = 14.3 ± 1.5 ms, N = 17; dlg mutant channel open time= 12.1 ± 0.7 ms, N = 42; P = 0.13).

All evidence strongly suggests that there are two subtypes of ionotropic glutamate receptors at the Drosophila NMJ: 1) receptors that are made up of the subunits GluRIIA+IIC+IID+IIE, and 2) receptors that consist of GluRIIB+IIC+IID+IIE. Our immunocytochemical results (Fig. 2) suggest that in dlg mutants, GluRIIB-containing receptors are selectively lost without any compensatory increase in GluRIIA-containing receptors. If this is true, the total number of glutamate receptors measurable electrophysiologically should decrease. To test this prediction, we measured the amplitude of glutamate-gated currents triggered by pressure ejection of 1 mM glutamate onto postsynaptic muscles. Figure 3C–D shows that, glutamate-gated currents were significantly smaller in dlg mutants, compared to controls (control = 1842 ± 255 pA at -60 mV, N = 10; dlg = 1044 ± 173 pA at -60 mV, N = 10; P = 0.018). Dividing by the single channel current amplitudes allows us to calculate the number of individual receptors opened. Control currents represent the opening of approximately 198 (1842/9.3) receptors. Currents in dlg mutants represent the opening of approximately 74 (1044/14.1) receptors. Since pressure ejection of glutamate onto embryonic muscles activates extrasynaptic as well as synaptic receptors, this decrease in electrophysiologically detectable glutamate receptors also demonstrates that the loss of immunocytochemically visible receptors shown in Fig. 2 is not due to dispersal of GluRIIB-containing receptors away from postsynaptic sites.

Recent studies suggest that Drosophila NMJ glutamate receptors are specifically localized opposite active zones, and that GluRIIA-containing receptors and GluRIIB-containing receptors are segregated from each other 263035. In other words, it is thought that individual postsynaptic densities (PSDs) contain either GluRIIA or GluRIIB, but not both. If this is true, then loss of one receptor subtype should cause some active zones to be without apposing receptor fields, while other active zones have relatively normal receptor fields. Our electrophysiological results (Fig. 3C–D) show that 63% (±12%) of all receptors are missing in dlg mutants. If GluRIIA and GluRIIB are segregated into different PSDs, and GluRIIB-containing receptors are selectively lost in dlg mutants, then 63% (±12%) of the individual synapses (active zone-PSD pairs) should be silent in dlg mutants. This should show up as a decrease in spontaneous excitatory synaptic current (sEJC) frequency, without a corresponding decrease in sEJC amplitude. sEJC frequency drops to 54% (±14%) of normal in dlg mutants (Figure 3E; control = 9.3 ± 1.6 Hz, N = 13; dlg = 5.0 ± 1.0 pA, N = 13; P = 0.03). sEJC amplitude in dlg mutants, however, is not significantly different compared to controls (Fig. 3F; control = 79 ± 7 pA, N = 13; dlg = 69 ± 9 pA, N = 13; P = 0.41). These results are consistent with selective loss of GluRIIB-containing receptors in NMJs where individual postsynaptic densities are composed exclusively of receptors containing GluRIIA or GluRIIB.

If postsynaptic glutamate receptor clusters are composed of receptors containing either GluRIIA or GluRIIB, and GluRIIB-containing receptors are selectively lost in DLG mutants, then there should be fewer postsynaptic glutamate receptor clusters per presynaptic active zone in dlg mutants. We tested this by triple staining the first instar NMJs with anti-HRP antibodies to visualize the presynaptic nerve, NC82, an antibody that marks presynaptic active zones 38, and anti-GluRIID antibodies 26, which label all postsynaptic glutamate receptors. The results are shown in Figure 4. NC82 and GluRIID immunoreactivity appear as distinct puncta where motor neurons contact postsynaptic muscle (Fig. 4A). In control larvae, each NC82 punctum is associated with a GluRIID punctum. Not every GluR cluster is associated with NC82 or HRP immunoreactivity, however, consistent with the previously-described presence of extrasynaptic glutamate receptors 394041. Thus, the ratio of postsynaptic glutamate receptor clusters to presynaptic active zones is greater than one (Fig. 4B). Specifically, control larvae show an average glutamate receptor cluster to active zone ratio of 1.33. In dlg mutants, however, this ratio is reduced to approximately one-half normal (Fig. 4B; control = 1.326 ± 0.15 GluR clusters/active zone, N = 312 GluR clusters from 5 animals; dlg = 0.64 ± 0.11 GluR clusters/active zone, N = 231 GluR clusters from 5 animals; P = 0.006). These results are consistent with a model in which: 1) GluRIIB-containing receptors are clustered independently of GluRIIA-containing receptors, 2) GluRIIB-containing receptor clusters are selectively lost in dlg mutants, and 3) selective loss of GluRIIB-containing receptors causes some presynaptic active zones to no longer be associated with postsynaptic glutamate receptor clusters.

Interestingly, some receptor clusters opposite active zones in dlg mutants were visibly less distinct (Fig. 4C, arrows), suggesting that GluRIIB-containing receptors are not only lost, but some receptors are slightly mislocalized.

Broadie & Bate 42 showed electrophysiologically that innervation triggers clustering and expression of functional glutamate receptors at the site of neuron-muscle contact. However, it has never been determined whether neuronal contact triggers localization of receptor protein, or local conversion of non-functional receptors to functional receptors. To answer this question, we repeated the critical experiments of Broadie & Bate but detected glutamate receptors immunocytochemically instead of electrophysiologically (Figures 5A–B).

In prospero null mutants, motor axon outgrowth is delayed and impaired, such that embryonic body wall muscles are variably innervated 42. Fig. 5A shows two neighboring hemisegments in a late stage (24 h AEL) prospero¹⁷ mutant embryo stained with anti-HRP to visualize motor axon terminals (green), and anti-GluRIIA antibodies (magenta). Ventral longitudinal muscles 12, 13, 6, and 7 (all unstained) are labelled in each hemisegment. The locations where the intersegmental nerve normally enter the ventral longitudinal muscle field in each hemisegment is marked with arrowheads. The muscles in the top hemisegment are innervated; ISNb (green) forms appropriate branches to each of the muscles, and GluRIIA clusters (magenta) are visible at the sites of innervation. As shown in Fig. 5A (top hemisegment, muscle 12), GluR clusters were typically visible even under growth cones, suggesting that receptors cluster at synaptic sites within minutes of nerve-muscle contact. In the neighboring uninnervated hemisegment (bottom of image), however, no GluRIIA clusters are visible. Thus, GluRIIA protein does not cluster in the absence of innervation. Similar results were obtained for GluRIIB (Fig. 5B). These results support the conclusion that postsynaptic glutamate receptors (containing either GluRIIA or GluRIIB) are clustered and upregulated only after contact by the presynaptic neuron. The signal between presynaptic neuron and postsynaptic muscle that triggers receptor clustering remains unknown.

Fig. 1 and previous studies 1617181943 show that DLG (like GluRs) is largely restricted to the postsynaptic region. We used prospero mutants to determine whether DLG localization also depends on contact by the presynaptic neuron. Figure 5B shows muscles 6 and 7 in a non-innervated prospero¹⁷ mutant hemisegment triple-stained for GluRIIB (green), DLG (magenta), and HRP (blue). As previously noted, glutamate receptor clusters are not visible in uninnervated muscles. Without innervation, DLG is also not localized; DLG remains dispersed throughout the muscle membrane in a pattern reminiscent of that seen when DLG is missing PDZ domains 1 and 2 13. In innervated muscles, DLG clusters appropriately at sites of muscle-nerve contact (Fig. 5C).

Because glutamate receptor clustering requires neuron-muscle contact, and DLG clustering requires neuron-muscle contact, it is possible that DLG clustering requires glutamate receptors. We tested this by visualizing DLG in homozygous Df(2L)SP22 mutant embryos. Df(2L)SP22 mutants contain a deletion that removes the genes encoding GluRIIA and GluRIIB, resulting in complete loss of NMJ glutamate receptors 262731. Figure 5C shows an innervated hemisegment from a homozygous Df(2L)SP22 embryo, stained with antibodies to the presynaptic vesicle protein CSP (green) and DLG (magenta). The innervation-dependent clustering of DLG does not depend on glutamate receptors; DLG clusters opposite presynaptic boutons even in homozygous Df(2L)SP22 mutants (Fig. 5C). Note, however, that some extrasynaptic DLG remains. Extrasynaptic DLG remains prominent until approximately 48–60 h after hatching (mid second instar stage; data not shown).

'Control' genotypes were either Oregon R (OR) or non-homozygous mutant siblings of the appropriate genotype. No statistically significant difference in any measurement was observed between OR and any other control genotype used in this study. Homozygous mutant embryos were identified through the use of an appropriate balancer chromosome expressing GFP. prospero[17] mutants are nulls that were a gift from Dr Chris Doe, University of Oregon. Df(2L)SP22 mutants remove both GluRIIA and GluRIIB, as previously described 31. dlg[X1–2] mutants 20 were gifts from Dr Vivian Budnik (University of Massachusetts Medical School).

Embryos and larvae were dissected and stained for immunocytochemistry and electrophysiology as previously described 43. When antibodies against any of the glutamate receptor subunits were used, preparations were fixed 30 min in Bouin's fixative. Otherwise, fixations were 30 min in 4% paraformaldehyde. Antibodies against GluRIIA (8B4D2, used at 1:100) 25 were produced from hybridoma cells and obtained from the University of Iowa Developmental Studies Hybridoma Bank (DSHB). Mouse NC82 antibodies were a gift from Erich Buchner and used at 1:100. Rabbit polyclonal GluRIIB antibodies 30 were used at 1:1000. Rabbit polyclonal anti-DLG antibodies 16 were used at 1:1000. Rabbit polyclonal GluRIID antibodies 26 were a gift from Stephan Sigrist and used at 1:500. All primary antibodies were visualized using fluorescently-conjugated (fluorescein, rhodamine, or CY-5) secondary antibodies (Jackson Immuno Labs, West Grove, PA) generated against the appropriate species (mouse or rabbit) and viewed using an Olympus FV-500 laser-scanning confocal microscope. Presynaptic terminals were visualized using fluorescently conjugated anti-HRP antibodies (Jackson Immuno Labs) directly conjugated to FITC, TRITC, or CY-5.

Receptor cluster sizes (Fig. 2C) were measured using an automated edge-finding/threshold-based macro run within NIH ImageJ software (v. 10.2 for OS X). Results using the automated procedure avoid experimenter bias and agree quantitatively with careful manual measurements 25. The 3D reconstruction of the portion of a larval NMJ shown in Fig. 1B was generated using Amira 3.1 (Mercury Computer Systems, Chelmsford, MA).

The number of glutamate receptor clusters per active zone (Fig. 4) was quantified as follows: first instar larvae of the appropriate genotype were dissected and triple-stained with antibodies against NC82, which marks presynaptic active zones, GluRIID, which marks all postsynaptic glutamate receptor clusters, and HRP to visualize the NMJ. These preparations were subsequently imaged using confocal microscopy. Z-projections from each image (which contained several NMJs and dozens of clusters) were split into the separate color channels using ImageJ, and the number of clusters in each channel (NC82 or GluRIID) was counted using ImageJ's particle analysis function. The number of GluRIID clusters in each image was then divided by the number of NC82 clusters in each image to calculate ratios that were then compared using a Student's T-test (Fig. 4B).

All electrophysiology (Fig. 3) was performed on ventral longitudinal muscle 6. whole-cell patch clamp measurements from embryonic muscles were performed as previously described 4143. Briefly, temporally and morphologically staged embryos were dechorionated in bleach, manually devitellinated and dissected, then treated with 1 mg/ml collagenase type IV (Sigma-Aldrich) for 60–90 s. Muscle 6 was whole-cell voltage clamped (-60 mV) in standard Drosophila embryonic saline using standard patch-clamp techniques. Data were acquired and subsequently analyzed using an Axopatch 1D amplifier and PClamp 9 (Axon Instruments, Union City CA).

Statistical significance in figures is represented as follows: *** = p < 0.001; ** = p < 0.01; * = p < 0.05. Unless otherwise specified (e.g. Fig. 2C, 3F), all statistical comparisons were made using unpaired T-tests, or (in the case of distributions) Kolmogorov-Smirnov tests. All error bars represent S.E.M.