Mapping of Multiple Neurotransmitter Receptor Subtypes and Distinct Protein Complexes to the Connectome

Summary

Neurons express various combinations of neurotransmitter receptor (NR) subunits and receive inputs from multiple neuron types expressing different neurotransmitters. Localizing NR subunits to specific synaptic inputs has been challenging. Here, we use epitope-tagged endogenous NR subunits, expansion light-sheet microscopy, and EM connectomics to molecularly characterize synapses in Drosophila. We show that in directionally selective motion-sensitive neurons, different multiple NRs elaborated a highly stereotyped molecular topography with NR localized to specific domains receiving cell-type specific inputs. Developmental studies suggested that NRs or complexes of them with other membrane proteins determine patterns of synaptic inputs. In support of this model, we identify a transmembrane protein associated selectively with a subset of spatially restricted synapses and demonstrate its requirement for synapse formation through genetic analysis. We propose that mechanisms that regulate the precise spatial distribution of NRs provide a molecular cartography specifying the patterns of synaptic connections onto dendrites.

eTOC

Sanfilippo et al. conditionally tag neurotransmitter receptor subunits using CRISPR-based modification of endogenous loci, expansion-light-sheet microscopy, immunopurification of associated proteins, and EM connectomics to map the diversity of synapses in the Drosophila brain. These studies reveal the complex molecular cartography of dendrites at the level of single-identified neuron types.

Introduction

Recent progress in electron microscopy (EM) has enabled the generation of single-synapse level connectomes of large brain volumes.^1–6 In Drosophila, EM reconstructions revealed extraordinary complexity and specificity of brain wiring. Hundreds of neuron types form specific patterns of connections with multiple partners in highly reproducible ways.^2,7 Single-cell sequencing has uncovered distinct patterns of mRNAs in different neurons for proteins involved in the development and function of synapses.^8,9 This is particularly striking in the transcriptional expression patterns of neurotransmitter receptor (NR) subunits.

Most NRs fall into three families: pentameric cys-loop ionotropic receptors, tetrameric glutamate ionotropic receptors, and G protein-coupled metabotropic receptors. In this study, we focus on NRs of the cys-loop superfamily. Cys-loop receptors mediate responses to the major excitatory and inhibitory neurotransmitters, acetylcholine and GABA, respectively, in the Drosophila brain. These subunits come together in different combinations to form ligand-gated ion channels. The expression of distinct combinations of NR subunits gives rise to multiple molecularly distinct receptor complexes.^10–14 Different mammalian GABA receptors of this family have been shown to localize to different domains of pyramidal cells (e.g., axon initial segment vs soma).^15,16 Previous studies in Drosophila have also shown that targeted expression of cDNAs encoding tagged versions of the GABA receptor subunit Rdl and the nicotinic subunit nAChRα7 resulted in localization to different dendritic domains in motor neurons¹⁷ and in T4/T5 neurons in the visual system.¹⁸ Voltage and calcium imaging in visual circuits in the fly revealed characteristic patterns of activity in different neuronal compartments.¹⁹ What underlies these subcellular specific activity patterns at the molecular level remains largely unknown. They likely include, in part, the subunit composition of NRs in the postsynaptic membrane and their spatial distribution in dendrites.

Localizing NRs at synapses in dendrites has been problematic for two reasons. First, it has been difficult to generate antibodies for multi-pass transmembrane proteins²⁰, which hinders specific immunohistochemical labeling of many NR subunits. To address this issue, we used CRISPR-modification to introduce various epitope tags into the endogenous NR loci. These modified loci allowed us to label NR subunits with highly specific commercially available antibodies while preserving their endogenous expression patterns. We also engineered conditional alleles for the selective labeling of tagged NR subunits in single neurons through cell-type-specific expression of recombinases. Second, the density of the neuropil and diffraction-limited light microscopy preclude localizing receptors to synapses in identified neurons. And third, there are substantial technical challenges of EM localization of proteins.²¹ To overcome these limitations, we take advantage of recent developments in Expansion Microscopy to localize protein in isotropically expanded tissue²² using custom-built lattice light-sheet^23,24 and commercially available light-sheet microscopy (ExLLSM and ExLSM, respectively), to achieve effective super-resolution.

Here, we localize tagged NR subunits in specific neuron types throughout the brain in adult and developing neurons at the level of single cells and at single synapses between identified neuron types at super-resolution. We focus on the distribution of seven different subunits in directionally selective motion-sensing neurons and demonstrate that these NRs are localized to specific spatial domains along the proximodistal dendritic axis. Using affinity-purification-mass-spectrometry, we identify a transmembrane protein implicated in synaptic adhesion associated with an NR subunit selectively localized to one of these domains. Our findings raise the possibility that NR protein complexes selectively localized to specific dendritic domains provide molecular cues specifying patterns of synaptic inputs from different neuron types.

Results

NR subunits are expressed in different brain regions

There are around 60 genes encoding subunits of NRs in the Drosophila genome and about 100 in the mouse and human genomes.^25–27 We generated tagged alleles of 11 NR subunits of the cys-loop superfamily, which form pentameric ligand-gated ion channels. These include receptors responsive to acetylcholine and γ-aminobutyric acid (GABA), the major excitatory and inhibitory neurotransmitters in the fly (Figure 1A; Table S1). In addition, we tagged the invertebrate-specific single cys-loop NR subunit that forms glutamatergic receptors. Unlike the subunits of the tetrameric iGluR family (e.g., AMPARs) found in vertebrates and invertebrates, which typically mediate excitation, these receptors gate chloride entry and thus are inhibitory.

Figure 1. Tagged NR subunits localize to synapses.

(A) Cys-loop superfamily NR subunits tagged in this study. For some NRs, two or three different tagged versions were made as indicated. Red asterisk, Lcch3, forms cationic channels with Grd and with CG8916 and chloride channels with Rdl.^11,14,59 The epitope tag was inserted into a poorly conserved region of the M3-M4 cytoplasmic loop in each NR subunit (Figures S1C and S1D). These proteins form homo- or heteropentameric channels.

(B) Whole animal tagging of NR subunits. For each conditionally tagged allele, a corresponding whole-animal tagged version was generated in which all cells expressing the receptor express a tagged version. Expression pattern of endogenously tagged alleles of nAChRα6 (top panel) and nAChRα7 (bottom panel). Dotted line, brain midline; yellow inset, neuropils of the central complex; cyan inset, optic glomeruli. Brp staining visualizes neuropils. Scale bars, 50 μm.

(C) Schematic for conditional tagging of NR subunits in sparsely distributed neurons. See Figure S4B for details. (D-F) Localization of NR subunits to domains of neurons with complex morphology in the central brain. MBONs have two main processes: one ending in a compact dendrite and another branched axonal process projecting to distinct brain areas. (D) Morphology of MBON14 or MBON13 – dendrites innervate select compartments of the mushroom body (yellow). Kenyon cells provide cholinergic inputs to MBON dendrites. Localization of nAChRβ1-smHA (E) or Rdl-smV5 (F) in MBON14 or MBON13. Arrowhead – dendrites; arrows – axonal projections; asterisk – putative axon initial segment. Scale bars, 25 μm.

(G) Schematic of synaptic active zone marked by Brp and the voltage-gated calcium channel Cacophony (Cac). In EM, cytoplasmic Brp protein localizes to a presynaptic T-bar structure associated with the presynaptic membrane (see panel K) and as a donut by STED and LLSM when immunostained for the Brp-directed antibody nc82.

(H) Brp with Cac at its center visualized using ExLLSM. The synapse shown is from the mushroom body calyx in the central brain. Lateral (left) and planar (right) views are shown.

(I) Schematic of NR subunits juxtaposing Brp and Cac.

(J) nAChRβ1 subunits cluster juxtaposed to an active zone in the calyx. Lateral views are shown.

(K) EM (right) and schematic (left) of multiple-contact synapse between projection neurons (PN) and Kenyon cells (KC) in the mushroom body (single synapse from EM data⁶⁰).

(L) Immunostaining of PN-KC synapses, as indicated.

(M) Examples of active zones and different NR subunits (as indicated) paired at synapses in the medulla neuropil.

Scale bars in H, J, L, and M panels represent unexpanded tissue size (adjusted for 4.65X expansion factor; see STAR Methods). See also Figures S1–S3 and Table S1.

Endogenous tagging of NRs is preferable to transgenic expression systems for studying NR localization, as the endogenous proteins preserve their unique, cell-type specific patterns and developmental expression.^18,28 Overexpression of tagged proteins can lead to their accumulation in the cell body and, in extreme cases, may result in the loss of their natural subcellular localization (Figures S1A and S1B). It is crucial to use endogenous tagging to study protein localization during development, as the developmental dynamics of transcriptional and post-transcriptional cell-type specific regulation are key determinants of protein localization.²⁹

To generate tagged alleles, the endogenous genomic loci encoding NR subunits were modified by directed knock-in of DNA sequences encoding epitope tags using CRISPR-targeted recombination^30,31 (Figure S1C). All tags were inserted into the cytoplasmic loop between the M3 and M4 transmembrane domains, which are shared amongst all cys-loop receptors (Figures 1A and S1D). Several studies have shown that NRs with insertions in this loop are functional and localized to postsynaptic sites.^32–35 NR subunits form homomeric or heteromeric NRs, or both (Figure 1A), resulting in a large array of distinct NR types with unique functional properties.^11,12,36

All tagged NRs localized to the neuropil (Figures 1B and S1E). Their expression patterns were similar to those seen with antibodies to the unmodified NR subunits (Figure S1F). Epitope-tagged proteins rescued lethality in cases where mutant alleles were not viable (see STAR Methods). Furthermore, all remaining homozygous tagged alleles were viable. Typically, the tagged NR subunits were broadly expressed throughout the brain, with some subunits highly enriched in specific neuropils. Each NR subunit exhibited a characteristic pattern of expression (Figures 1B and S1E–S1F; also see Figure S4A). Tagged proteins were expressed at levels similar to the untagged version as quantified by staining with different concentrations of antibody to the native proteins and normalizing to the intensity of staining to the pan-synaptic protein Brp (Figure S1G).

Due to the density of processes within neuropils, it is not possible to assign the expression of proteins detected by immunofluorescence to specific neuron types. To overcome this limitation, we generated inducible alleles and a new approach to tag NR subunits in identified cell types with single-cell resolution (Figures 1C, S1C and S4B). Reagents to label virtually any neuron in Drosophila in this way are available.^37,38 In this study, we explored the distribution of NR subunits in different neurons in the visual system and the mushroom body.

The mushroom body is the center for associative learning in insect brains, and its sensory inputs and output synapses are cholinergic^39,40 (Figure S3A). As expected from RNA-seq data⁴¹, we observed the nAChRβ1 subunit in the dendrites of two different MBONs (Figures 1D and 1E). In contrast, the subcellular distribution of the GABAergic subunit Rdl in MBONs was highly cell-type-specific (Figures 1D, 1F and S3C–S3D). For instance, in MBON05, Rdl localized to both axons and dendrites, while it was predominantly restricted to dendrites in MBON11 (Figures S3C and S3D). Thus, tagged NR subunits localize to specific neuronal domains, which may even differ between neurons with closely related functions. We next sought to address whether the tagged NR subunits were localized to synapses.

Tagged NR subunits localize to synapses

The resolution of light microscopy is not sufficient to localize NRs to synapses. To increase the effective resolution, we imaged expanded tissue²² using a lattice light-sheet microscope²⁴ (ExLLSM). This method allows imaging of large volumes at an effective resolution of approximately 60 by 60 by 90 nm.²³ We assessed the localization of the tagged NRs by comparing images of synapses in the mushroom body with those previously obtained using stimulated emission depletion microscopy (STED) and EM.

Previous STED studies of the mushroom body (MB) calyx identified presynaptic sites characterized by Brp-stained donut-shaped structures with the Ca2+ channel Cacophony (Cac) at the center⁴² (Figure 1G). Brains bearing an ALFA-tagged allele of Cac and stained for both the ALFA tag and Brp were imaged. We observed presynaptic structures in the MB calyx and in the medulla region of the optic lobe of similar dimensions to those previously described in the MB calyx in STED and EM studies (Figures 1G–1J and S2A). Tagged nAChRβ1 clustered in juxtaposition to these presynaptic sites (Figures 1I–1J and S2B). The synaptic structures identified by ExLLSM for other NR subunits in the calyx and medulla were also similar (Figures 1K–1M and S2C–S2E). All NR subunits were preferentially associated with Brp (see Figures S3M–S3P for quantification), consistent with their synaptic localization. In addition, tagged nAChR subunits localized in different types of synaptic structures similar to those described by EM^39,42,43 (Figures 1K, S3A–S3B and S3E–S3L).

In summary, endogenously tagged NRs visualized by ExLLSM were distributed in NR-type-specific patterns within the CNS and selectively localized to postsynaptic sites, which is consistent with previous studies using EM and STED.

NR subunit distribution matches the EM connectome

We next sought to determine whether the patterns of NRs correlated with the distribution of inputs from cholinergic, GABAergic, and glutamatergic neurons as determined by connectomic and gene expression studies. We focused on circuits in the fly visual system and, in particular, the medulla neuropil. The repetitive pattern allowed us to look at many different neurons of the same type in a single animal, and it is straightforward to generate sparsely labeled neurons of the same type. In the medulla neuropil, overlapping processes of many different neuron types form stereotyped circuits (Figure 2A). Each neuron type expresses different levels and combinations of transcripts encoding NR subunits⁸ (Figure S4A).

Figure 2. NR subunit distribution matches the EM connectome.

(A) Morphologies of a small selection of neuron types of the ~150 different types in the fly optic lobe (Adapted from ref.⁶¹). Three of ten medulla neuropil layers are labeled (M1, M5, and M10).

(B) NR subunit expression (green) across medulla layers (M1-M10). Presynaptic marker Brp (magenta) is detected with anti-Brp antibody throughout all medulla layers. NR subunits are detected with epitope-tag-specific antibodies (anti-V5, anti-Ollas, anti-HA, anti-ALFA). Scale bar, 10 μm.

(C-E) Localization of GluClα-smV5 (C), nAChRβ1-smHA (D), and Rdl-smV5 (E) in dendrites of L5 neurons. Left, EM-based reconstruction of L5 dendrites and annotated synapses for each neurotransmitter type⁷. Right two panels, neuron morphology (gray) and NR subunits (color-coded). Scale bars, 5 μm.

(F) Quantification of data shown in E-G and from the EM⁷ (EM, n=7; Rdl, n=15; GluClα, n=12; and nAChRβ1, n=8). Bonferroni adjusted p-value (*), < 0.05, (***), < 0.001 from Wilcoxon rank-sum test. See text for a discussion of discrepancies between EM and tagged puncta.

(G) Distribution of GluClα-smV5 in T4 dendrites. Quantification is the same as in Figure S5F. Scale bars, 5 μm.

See also Figure S4.

The ten layers of the medulla comprise the highly branched processes and synapses of many neurons (>10,000 neurons, >100 neuron types, and ~2 million synapses).⁷ Some receptors are broadly expressed, whereas others are preferentially enriched in specific layers (Figure 2B). For instance, the inhibitory GABA receptor subunit Rdl, which can form homomeric and heteromeric receptors^11,14, is expressed in most layers of the medulla. By contrast, the expression of Lcch3, another GABA receptor subunit that can form heteromeric receptors with Rdl¹⁴, is more restricted (Figure 2B). Differential localization was also observed for different nAChR subunits (Figure 2B).

EM-level connectomes have established that many neuron types in the medulla receive inputs from multiple types of presynaptic neurons.^7,44 For instance, GluClα localized to domains of L5, which predominantly receive glutamatergic inputs in medulla layers M1 and M5 (Figure 2C). The number of puncta seen for GluClα in L5 neurons correlated well with EM data (Figure 2F). These neurons also receive cholinergic and GABAergic innervation to the same and different dendrites in discrete and reproducible patterns, and this is reflected in the distribution of Rdl and nAChRβ1 (Figures 2D–2F). Discrepancies between synapse numbers were observed in some cases with Rdl and different acetylcholine receptors. This may represent NRs that are extrasynaptic (e.g., small Rdl puncta, see Figure S2E) and different receptors for the same neurotransmitter at different synapses in the same neuron (e.g., multiple nAChRs). In general, however, there was a good correlation between the distribution of receptor puncta and synapses determined by EM across several neuron types (Figures 2G, S4C–S4F and S5F).

NR subunits are differentially localized along dendrites

The precise distribution of synaptic inputs from different neurons along the proximodistal axis of T4 dendrites is proposed to play a crucial role in motion detection. T4 dendrites receive inputs from GABAergic, cholinergic, and glutamatergic neurons from eight identified cell types in specific domains along the proximodistal axis^4,44 (Figure 3A). The synaptic inputs to T5 are different (Figure S5A; see below), with different neuron types also forming synapses within specific dendritic domains⁴.

Figure 3. NR subunits are differentially localized along dendrites.

(A) Upper panel, EM reconstruction of different presynaptic inputs, as indicated, along the proximodistal axis of T4 dendrites. Dendrites span an average of three columns. Lower panel, annotated synapses for different neurotransmitter inputs are shown. Colored arrows in the lower panel show domains targeted by color-matched presynaptic inputs from the upper panel. TmY15 is an exception forming synapses across the length of T4 dendrites.

(B-D) Localization of tagged NR subunits in dendrites of single T4 neurons. Left-most column, EM reconstruction of glutamatergic, GABAerbic, and cholinergic inputs. Central columns, NR subunit pattern observed by conditional tagging in single T4 dendrites. Right-most column, subset of EM-annotated synapses as shown in the lower part of panel A. Scale bars, 5 μm.

(E) Quantification of NR subunit expression along the normalized proximo-distal axis of T4 dendrites (see STAR Methods). Rdl, n = 6; CG8916, n=4; Lcch3, n=3; GluClα, n = 4; nAChRα1, n = 5; nAChRα5, n = 5; nAChRβ1, n = 10.

See also Figure S5.

The distribution of NRs in T4 dendrites was consistent with the neurotransmitters used by the synaptic inputs. GluClα, as described above, was highly localized to the distal domains (Figures 3B and 3E). By contrast, Rdl was strongly enriched in the proximal region, with additional puncta sparsely scattered throughout the arbors (Figures 3C and 3E). T4 neurons also express two additional GABA NR subunits, Lcch3 and CG8916 (see Figure S4A). These two subunits were not localized proximally but exhibited a common distribution with enrichment in the distal region and then sparsely throughout the rest of the dendrite (Figures 3C and 3E). The proximal enrichment of Rdl matches the innervation pattern by three different GABAergic neuron types (i.e. Mi4, C3, and CT1). By contrast, TmY15 synapses scattered along the length of T4 dendrites may signal through NRs containing Lcch3 and CG8916.

Different nAChR subunits also show different distributions in T4 dendrites. nAChRα5 localized to the middle domain of T4 dendrites, whereas both nAChRβ1 and nAChRα1 localized to the distal domain (Figures 3D and 3E). These patterns correlate with the cholinergic inputs from overlapping Mi1 and Tm3 neurons in the middle domain and to dendrodendritic synapses of overlapping T4 dendrites in the distal domain. These data suggest that the subunit composition of NRs to the same neurotransmitter is unique to specific synaptic inputs.

T5 dendrites receive inputs different from T4, and these also showed specificity along the proximodistal axis (Figures S5A–S5E). Rdl, nAChRα5, nAChRβ1, and nAChRα1 subunits were distributed similarly to T4 (Figures S5C–S5E). GluClα was not detected in T5 dendrites, consistent with the absence of glutamatergic inputs (Figures S5A and S5B). CG8916 and Lcch3 also lacked the enrichment to the tips observed for T4 dendrites (Figures S5C and S5E). There were twice the number of nAChRβ1 puncta in T5 as in T4 dendrites (Figure S5F). This is consistent with the additional cholinergic input T5 dendrites receive in the distal region by Tm9 neurons (Figures S5A and S5D). Three other Tm neuron types (Tm1, Tm2, and Tm4) provide inputs to the middle domain in a pattern similar to the distribution of nAChRα5 (Figures S5A and S5D).

The distribution of nAChRα1, nAChRβ1, and nAChRα5 partially overlap in the middle domain. To assess whether these nAChR subunits were in the same or different synapses, we tagged nAChRα1 or nAChRα5 in combination with nAChRβ1 in the same neuron with different epitopes (Figure S5G). nAChRα1 puncta largely co-localized with nAChRβ1, suggesting that these are found at the same synapses (Figures S5H and S5I). By contrast, nAChRα5 and nAChRβ1 did not co-localize, demonstrating that these are not at the same synapse (Figures S5H and S5I). This is consistent with findings showing that nAChRα5 subunits, which are closely related to the vertebrate α7 nAChR subunit known to form homomeric receptors^45,46, can also form homomeric receptors¹². In contrast, there is evidence that nAChRα1 can only assemble into heteromeric receptors, which must include β nAChR subunits.¹⁰ Thus, even within the same dendritic domain, different cholinergic neuron types may form synapses selectively onto postsynaptic sites expressing different receptors.

In summary, the distribution of NR subunits reflected the specific arrangement of various GABAergic, cholinergic, and glutamatergic inputs onto T4 or T5 dendrites with spatial specificity along their proximodistal axis. Different neuron types, which use the same neurotransmitter, may communicate through molecularly distinct receptors within the same or different dendritic domains.

Localization of NR subunits during development is cell-type specific

We next sought to explore how NRs become localized to synapses during development. NRs could localize directly to the discrete domains where synapses form. Alternatively, NRs could be initially uniformly distributed and subsequently stabilized at synapses, downregulated in incorrect locations, or both. Analysis of whole-animal tagged NR subunits for Rdl-smV5, nAChRβ1-smHA, and GluClα-smV5 showed progressive accumulation of NR subunits in the developing medulla neuropil (Figures S6A–S6D). To address whether this accumulation occurs similarly in different neurons, we turned to single-cell analysis in T4, T5, and L5 medulla neurons, where these NR subunits are expressed throughout pupal development (Figures 4 and S6A).

Figure 4. Localization of NR subunits during development is cell-type specific Time course of NR subunit accumulation during development of T4.

(A, C), T5 (B, D) and L5 dendrites (E, F) of Rdl-smV5 (A-B, E) or GluClα-smV5 (C-D, F). Times shown are 48h and 72h after pupal formation (APF) and adult (Ad). Schematic of dendrite development with NR distribution shown for these time points. For T4 and T5, quantification of puncta distribution along the normalized proximodistal axis at 72h APF and adult are shown below the schematics (T4: adult quantification data same as Figure 3E; Rdl 72h APF, n = 4; GluClα 72h APF, n= 6. T5: adult quantification data same as Figure S5E, Rdl 72h APF, n = 7; GluClα 72h APF n= 2). Scale bars, 5 μm.

See also Figure S6.

In sparsely labeled T4 and T5 neurons, Rdl accumulated as large puncta at the proximal region of dendrites at 48h APF, corresponding to the onset of dendrite extension (Figures 4A and 4B). As dendrites extend, Rdl remains enriched proximally with small fainter puncta distributed along their length. At 48 hrs, GluClα and nAChRβ1 puncta are also seen in incipient T4/T5 dendrites, but, over time, disappear from the proximal region of the dendrite and accumulate more distally (Figures 4C–4D and S6E–S6F). This is particularly striking for GluClα. It accumulates throughout T4 and T5 dendrites at 48 and 72 hrs APF, disappears from T5 dendrites (consistent with the lack of glutamatergic inputs to T5) between 72 hrs APF and eclosion, and is retained only in the distal domain of T4 dendrites in the adult (Figures 4C and 4D).

The accumulation of NR subunits in developing L5 neurons was different. In contrast to T4, Rdl puncta were observed throughout L5 terminals in the medulla from 48 hrs through 72 hrs APF but then disappeared from all but a highly restricted domain in medulla layers M1/M2 (Figure 4E). Also, in contrast to T4/T5 neurons, GluClα localized to terminals early, by 48 hrs APF, and this pattern remained into the adult (Figure 4F). To assess the timing of GluClα localization to developing synapses, we imaged GluClα-smV5 in developing L5 neurons with super-resolution using ExLSM. The analysis revealed progressive juxtaposition of GluClα puncta with Brp, as neurons develop from 48h APF to eclosion (Figures S6G–S6I); most GluClα puncta early in development do not cluster with Brp (see arrowheads, Figure S6G). Altogether, different NR subunits localize to dendrites with distinct developmental dynamics and can be cell-type-specific rather than an intrinsic property of a given NR.

A fly homolog of the synaptic adhesion protein ADAM22 forms a complex with GluClα

The exquisite localization of NR subunits to synapses in distinct dendritic domains raised the possibility that proteins associated with different NRs may regulate their localization, contribute to synaptic specificity, or both. As a first step to assessing this possibility, we focus on GluClα. Using an affinity purification mass-spectrometry (AP-MS) workflow, proteins associated with epitope-tagged versions of GluClα were identified in extracts of fly brains. Two different tagged versions of GluClα and two complementary sets of controls were used (Figure 5A). Mind-meld (Mmd), a single-pass transmembrane protein, was specifically bound with both versions of GluClα (Table S2). Co-IP experiments confirmed that Mmd interacts with GluClα in homogenates of brain tissue but not with Rdl (Figure 5B).

Figure 5. Mmd forms a complex with GluClα.

(A) Design of AP-MS experiment used to identify interactors of GluClα. Tagged proteins were purified as indicated from head homogenates and subjected to mass spectrometry. Under these conditions, a single protein, Mmd, co-purified with both tagged versions of GluClα but not in controls (see STAR Methods and Table S2). All samples represent three biological replicates.

(B) Mmd and GluClα form a complex. Immunoprecipitation experiment. Left panel, extracts from fly heads carrying smGFP-smV5-tagged GluClα and smGFP-smV5-tagged Rdl probed with anti-V5 and anti-Mmd antibodies. Right panel, a nanobody against GFP was used to pull down tagged GluClα or tagged Rdl and probed on blots as indicated. GluClα forms a complex with predominantly one form of Mmd. Asterisk, weak band co-precipitating with tagged Rdl.

(C) GluClα-smV5 staining is lost in an mmd mutant. Scale bar, 10 μm.

(D) Models. Bottom, synaptic complex comprising ADAM22, the mouse homolog of Mmd. ADAM22 on the postsynaptic membrane and ADAM22 or a related protein on the presynaptic membrane are bridged by interactions with a dimer of LGI1⁵⁴; Top, localization of GluClα and Mmd at a synapse in Drosophila as shown in this study. Interactions of Mmd with other proteins are not known.

See also Table S2.

As examined by confocal microscopy, there was a complete overlap in the distribution of Mmd and GluClα (Figure 5C). To assess the role of Mmd in the formation of synapses containing GluClα, we analyzed brains mutant for mmd. Strong loss of mmd resulted in a marked reduction in anti-GluClα staining (Figure 5C). Mmd is homologous to mammalian ADAM22, which has been shown to form a complex with AMPA receptors and acts as a synaptic adhesion molecule through its binding to LGI1⁴⁷ (Figure 5D). These data raise the possibility that in T4 neurons, GluClα and Mmd form a complex selectively in the distal domain of their dendrites, which, in turn, specifies the pattern of Mi9 inputs. We next sought to directly visualize GluClα and Mmd at these synapses.

Mi9 axons form synapses juxtaposing GluClα and Mmd in distal T4 dendrites

EM studies revealed that seven to nine Mi9 axon terminals, organized into retinotopic columns, are evenly distributed across the proximal, central, and distal domains of T4 dendrites (Figure 6A). Mi9 neurons synapse, however, only onto the distal-most domains⁴ (Figure 6A). Other neuron types form synapses within different domains along the proximodistal axis of these dendrites.⁴ Mi9 is the only glutamatergic input to T4 dendrites. Here, we applied ExLSM to achieve sufficient resolution to visualize synapses between Mi9 and T4 dendrites and localize proteins within these synapses.

Figure 6. Mi9 axons form synapses juxtaposing GluClα and Mmd in distal T4 dendrites.

(A) Identification of synapses between identified neuron types using ExLSM. Synaptic partners, and pre- and postsynaptic proteins were visualized using four-channel light-sheet microscopy on expanded tissue (see STAR Methods). Colors correspond to expression in Panels B, and C. Arrow denotes proximo-distal axis along T4 dendrites.

(B-C) Mi9 forms synapses selectively in the distal domain of T4 dendrites. (B) T4 dendrites project across multiple columns. Mi9 axon terminals, one per column, contact all domains of T4 dendrites but only form synapses distally. (C) Upper row, synapses form between T4 and Mi9 in the distal domain (red box in B). Lower row, centrally located Mi9 terminals do not form synapses (yellow box in B).

(D) GluClα co-localizes with Mmd at multi-contact synapses. Arrowheads point to multiple postsynaptic sites at a single multi-contact synapse co-labeled with Mmd and GluClα.

(E) Mmd antibody recognizes an epitope in the Mmd ectodomain, and the V5 antibody labels a V5 tag in the cytoplasmic tail of Mmd knocked into the genomic locus.

(F) GluClα is tagged in a single T4 neuron. All other T4 synapses express Mmd but not tagged GluClα. Arrow points to a single postsynaptic site of a multiple-contact synapse co-labeled with Mmd and GluClα. Scale bars, 200 nm, represent unexpanded tissue size (adjusted for 4.65X expansion factor; see STAR Methods).

See also Figure S7.

To detect synapses between Mi9 and T4, we coupled labeling of GluClα in single T4 dendrites with staining of the membranes of Mi9 axon terminals and presynaptic Brp within them (Figures 6A and 6B). Control experiments verified our ability to assign Brp to individual Mi9 terminals and separate them from Brp in other processes in the neuropil (see STAR Methods, Figures S7A and S7B). Consistent with EM studies, presynaptic sites in Mi9 terminals contacting the distal region of T4 dendrites were associated with GluClα expressed in these dendrites, while presynaptic sites of Mi9 terminals overlapping with the middle and proximal domains were not (Figures 6C and S7C–S7E). We next sought to determine whether Mmd also localized with GluClα at synapses between Mi9 and T4 dendrites.

We first confirmed via ExLSM the localization of Mmd and GluClα at synapses using both an antibody to the extracellular domain of Mmd and an antibody to an epitope tag inserted into its cytoplasmic domain (Figures 6D and 6E). To assess the localization of Mmd with GluClα in T4 neurons, we labeled single T4 neurons in which GluClα was selectively tagged with smV5 and co-stained with an anti-Mmd antibody. Each GluClα punctum co-localized with Mmd (Figure 6F). Mmd staining extended beyond anti-GluClα immunoreactivity. This was expected, given that each Mi9 presynaptic site contacts multiple postsynaptic elements from other T4 neurons, and only a single postsynaptic neuron was labeled with tagged GluClα at these synapses.⁴ In summary, Mmd localized to the same synapses as GluClα in the distal domain of T4 dendrites juxtaposing Mi9 terminals.

Discussion

In this study, we described a strategy to study single synapses between identified cell types by combining genetic tools, EM-based connectomics, and protein localization through expansion microscopy. To devise probes for specific synapses, we developed tools to tag and map the distribution of endogenous NR subunits of receptors for the major neurotransmitters in Drosophila in single neurons and at single synapses between neurons. These studies highlighted the complexity of different NR subunit distributions in dendrites, the developmental dynamics of NR accumulation at different synapses, and their association with other proteins at subsets of synapses. These tagged receptors provide a resource for studies focusing on the subcellular localization of neuronal proteins, the assembly of synapses, and synaptic plasticity. As we discuss below, the unique features of wiring specificity of T4 and T5 dendrites suggest that in some developmental contexts, targeting NRs to discrete dendritic domains provides a spatial map of molecular determinants controlling synaptic specificity.

Mapping NR subunits to specific synapses in single neurons

Our approach to conditionally tag receptors by modifying genomic loci preserves their endogenous mRNA expression levels, patterns, and temporal dynamics. We chose to insert epitope tags into poorly conserved and unstructured regions of variable length within a large cytoplasmic loop found in all cys-loop receptors. The distribution of the tagged NRs accurately reflects the localization of endogenous receptors. Tagged NRs selectively localized at sites adjacent to presynaptic partners (i.e. Brp and Cac) when visualized by either ExLLSM or ExLSM, and the distribution of receptors matched the neurotransmitter specificity of the neurons providing synaptic inputs. The increasing availability of genomes for comparative sequence analysis and the development of structure prediction tools such as AlphaFold⁴⁸ facilitate the identification of sites to successfully tag other classes of NRs and other synaptic proteins. The extensive synaptic connectivity maps in flies, reagents for manipulating specific cell types, and the array of endogenously tagged NR subunits reported here provide an opportunity to characterize the distribution of these receptors in many different circuits regulating a broad range of brain functions and behaviors.

For some NRs, receptor puncta and synaptic inputs identified at the EM level were concordant (i.e. GluClα). For other NRs (e.g. Rdl and nAChRs NR subunits), there was considerable variation between receptor puncta and EM. In some cases, for instance, more Rdl puncta were seen than predicted from EM. This may reflect extrasynaptic receptors or receptor populations within trafficking complexes en route to synapses. By contrast, there were fewer puncta for some nAChR subunits than predicted from EM, and this may reflect mutually exclusive expression of NR subunits in different populations of synapses receiving inputs from different cholinergic neuron types. This is likely to account for the discrepancy between cholinergic inputs inferred from the EM and puncta for single nAChR subunits in T4 and T5 dendrites. Matching different NR subunits to specific presynaptic inputs, as we showed for Mi9 and GluClα, provides a rigorous way to compare EM-based connectomics with light level analysis.

A lack of postsynaptic markers has hindered our understanding of the developmental dynamics of synapse formation in the fly brain. Tagged NR receptors have provided a key set of reagents for exploring this process. We demonstrate that the timing of the localization of different NR subunits to specific dendritic domains is specific to different neuron types. This may result from cell-autonomous mechanisms that direct receptors and associated factors, such as Mmd, to specific neuronal domains. Alternatively, the interaction with presynaptic neuron membrane proteins could influence the timing of clustering of NR subunits in the postsynaptic cell. Mechanisms that regulate the trafficking of receptors to the plasma membrane could also gate synapse formation in time and space, for instance, by differential internalization of GluClα receptors in T4 and T5 dendrites. Our method of conditionally tagging receptors by modifying the endogenous locus maintains the endogenous mRNA expression levels, patterns, and temporal dynamics of these receptors. With the super-resolution provided by ExLSM, the tagged NR subunits we present are an invaluable resource for investigating the dynamics of synapse formation at fly central synapses.

Different receptors exhibit discrete localization patterns. For instance, the distribution of seven different tagged NR subunits for glutamate, GABA, and acetylcholine in the cys-loop superfamily in direction-sensitive T4 neurons was particularly striking. The stereotyped patterns of each class of receptor matched the pattern of the neurotransmitters used by their respective presynaptic inputs. Unexpectedly, different NRs to the same neurotransmitter are also localized to different domains, receiving input from different presynaptic partners. This was seen for both cholinergic (excitatory) and GABAergic (inhibitory) synapses. The differential distribution of NR subunits to different spatial domains may contribute to the unique computational features of T4 dendrites.⁴⁹ How these domains are established in neurons with diverse and often complex morphologies is not known. Perhaps the segregation of proteins to different domains may, at a mechanistic level, share features in common with the establishment of cell polarity domains in other cell types, including apical basal polarity in epithelial cells⁵⁰.

Cys-loop GABA receptors that differ in NR subunit composition are found in distinct domains in pyramidal cells of the mouse cortex.¹⁶ It is likely that the complexity of receptor distributions in mammals also extends to different domains within the same dendrites as we have described here. In the mammalian brain, there is a great diversity of the cys-loop family of GABA receptors (i.e. GABAAR), with 19 distinct genes encoding GABAAR subunits.¹⁶ Thus, tagging approaches in the mouse similar to what we report here in the fly may provide a way to uncover the spatial distributions of NR subunits and combinations of them at different synapses. As these receptors have different physiological properties, their spatial distributions may contribute to understanding information processing in the dendrites of mammalian neurons.

NR subunit diversity and synaptic specificity

Our analysis of T4 neurons raised the intriguing possibility that in some developmental contexts, neurotransmitter receptors may serve as recognition molecules specifying the pattern of presynaptic inputs. In these dendrites, each presynaptic neuron type forms synapses within restricted domains along the proximodistal axis⁴⁴, and these patterns correspond to the distribution of different NR subunits. Mi9 neurons, for instance, only form synapses within the distal domains of T4 dendrites, while closely related Mi1 and Mi4 neurons form synapses in the middle and proximal regions, respectively. This specificity is particularly striking as the dendritic arbors of many T4s overlap extensively. Each input axon terminal contacts the entire range of dendritic domains and yet makes synapses only within restricted spatial domains (Figure 7). A simple model to account for this specificity is that targeting NR subunits to specific dendritic domains provides a spatial map of molecular signposts recognized by cell surface recognition proteins selectively expressed on the surface of different presynaptic terminal arbors. Alternatively, specific molecular determinants may be arranged along the proximodistal axis in discrete dendritic domains, and these may recruit both synaptic inputs and subsets of NR subunits to these sites. In either case, our studies suggest a close relationship between the targeting of NR subunits to different spatial domains and the specificity of synaptic inputs.

Figure 7. Model: NR subunits as synaptic specificity determinants in dendrites.

Neurons providing presynaptic inputs to T4 dendrites in each column in the medulla encounter the full range of spatial domains due to the staggered overlap of many T4 dendrites from adjacent columns. These inputs, however, discriminate between different T4 dendrites and selectively target specific dendritic domains. Based on this study, we propose that in some contexts, NRs or proteins in complexes with them are sorted to specific dendritic spatial domains and serve as determinants of synaptic specificity.

(A) Side and top views of 20 T4 dendrites contacting a single column. Upper panel (side view): The terminals of a single Mi9 and Mi1 in the central column (column 1) are depicted. Lower panel (top view): Dotted circular lines outline each medulla column. A single T4 dendrite is highlighted in red. Mi1 and Mi9 neurons in columns surrounding column 1 are shown in a slightly weaker shading than in column 1. The patterns of synaptic inputs to different T4s from Mi9 and Mi1 neurons in column 1 are shown in panels B and C. EM data in A-C from ref.⁷.

(B-C) All T4 dendrites receive inputs from both Mi9 and Mi1 neurons. In each column, different subsets of T4 dendrites receive input from a single Mi9 (B) or a single Mi1 (C) neuron. We describe one column as an example (column 1). Each circle demarcated with dotted lines indicates a different column; each column contains a single Mi9 and a single Mi1 axon terminal. Cyan indicates synapses established selectively by Mi9 in the distal domains of T4 dendrites (B), and yellow indicates synapses formed by Mi1 in the same column selectively with the central domains of a different subset of T4 dendrites (C). Synapses formed by Mi9 and Mi1 in the surrounding adjacent medulla columns are not shown. The right panels show the individual T4 neurons with distal domains overlapping (B) and overlapping middle domains (C). Colored arrows matching the dendrite colors indicate the proximal (PROX.) to distal (DIST.) polarity of these dendrites. The large black dots indicate the base of the dendrite in the proximal domain.

(D) Summary illustrating the distribution of NR subunits across different domains of the T4 dendrites. The pattern of innervation of GABAergic, cholinergic, and glutamatergic input neurons matches the distribution of NR subunits (see Figure 3). NR subunits may serve as specificity determinants, allowing presynaptic neurons to distinguish between different T4 dendritic domains. Alternatively, other spatial determinants, such as Mmd in T4, may recruit both NRs and presynaptic inputs to the same domain.

This model is consistent with studies in mammals. NRs have been shown to associate with other postsynaptic proteins that promote adhesion between pre, and postsynaptic membranes, and these complexes may be selectively localized.^51,52 Here, for instance, we report the identification of a transmembrane protein, Mmd, in the distal domains of T4 dendrites in close association with the GluClα receptor. A mammalian homolog of Mmd, Adam22, also co-localizes to synaptic glutamate receptors, albeit of a different class, and promotes adhesion between presynaptic and postsynaptic membranes^47,53,54 (Figure 5D). Alternatively, there is evidence that the extracellular domains of NR subunits can directly interact with proteins on the presynaptic membrane. For instance, the extracellular domain of the iGluR subunit GluA1 can interact with presynaptic neuronal pentraxin receptors, and this interaction can support the formation of synapses in a heterologous system.⁵⁵ Similarly, the N terminal of the α1 GABAAR subunit has been shown to interact with neurexin-2β and modulate GABAAR function.⁵⁶ Together, these observations raise the notion that, in some cases, the NR subunits are intimately involved in matching pre and postsynaptic membranes.

Perspective

The complexity of neural circuit structure has become increasingly clear with the completion of dense EM connectomes.^1–5 Extensive studies have argued that different domains along the proximo-distal axis of T4 dendrites play a crucial role in direction-specific motion processing.^49,57,58 The identification of different NRs and combinations of them within different domains provides an opportunity to understand computations at the molecular level.

How these specific molecular domains in dendrites form and their patterns of synaptic inputs emerge during development remains enigmatic. Localization of NR subunits to different domains at early stages of dendrite development raises the possibility that this molecular diversity contributes to determining the spatial distribution of specific synapses in dendrites. The use of tagged NRs and cell type-specific manipulation with EM-based connectomics, ExLSM, genetics, and biochemical methods provides a way of understanding how NRs become localized to specific domains and linking these to the specificity of synaptic inputs.

STAR Methods

RESOURCE AVAILABILITY

Lead contact

All requests for additional information and reagents should be directed to and will be fulfilled by the lead contact, S. Lawrence Zipursky (lzipursky@mednet.ucla.edu).

Materials availability

Flies generated in this study have been deposited to the Bloomington Drosophila Stock Center. Plasmids have either been deposited to Addgene or are available upon request.

Data and Code Availability

All data reported in this paper will be shared by the lead contact upon request. This paper does not report original code. Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

EXPERIMENTAL MODEL AND SUBJECT DETAILS

Fly husbandry

Drosophila melanogaster was reared on cornmeal/molasses medium at 25°C in a humidity-controlled incubator. Females were dissected for experiments unless otherwise noted. White prepupae were collected for developmental studies and designated as 0h after pupal formation (APF). Stocks used and generated in this study are listed in the key resources table. Genotypes used in each figure panel and related immunofluorescence staining conditions are reported in Table S3. Transgenic flies were generated by integrating DNA constructs into specific landing sites, while targeted alleles were created through CRISPR-mediated homologous recombination, as described below. This was carried out using a commercial injection service (BestGene, Inc.).

KEY RESOURCES TABLE.

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
rabbit anti-nAChRα6	Ref.73	N/A
rat anti-nAChRα7	Ref.76	N/A
rabbit anti- Rdl	Ref.77	N/A
rabbit anti-MdGluClα	Ref.78	N/A
rabbit anti-Mmd	Guo et al., manuscript under preparation	N/A
rabbit anti-Ollas	GenScript	Cat# A01658
rat anti-Ollas (L2)	Novus Biologicals	Cat# NBP1-06713, RRID:AB_1625979
rat anti-HA (3F10)	Roche	Cat# 11867423001, RRID:AB_390918
mouse IgG2a anti-V5 (SV5-Pk1)	abcam	Cat# ab27671, RRID:AB_471093
chicken anti-V5	abcam	Cat# ab9113, RRID:AB_307022
mouse anti-V5-Tag:DyLight®550	Bio-Rad	Cat# MCA1360D550GA, RRID:AB_2687576
FluoTag X2 anti-ALFA::Atto488	NanoTag Biotechnologies	Cat# N1502-At488
FluoTag X2 anti-ALFA::Alexa Fluor 568	NanoTag Biotechnologies	Cat# N1502-AF568
FluoTag X4 anti-RFP::AZDye568	NanoTag Biotechnologies	Cat# N0404-AF568
rabbit anti-DsRed	Takara Bio	Cat# 632496, RRID:AB_10013483
guinea pig anti-RFP	Synaptic Systems	Cat# 390004, RRID:AB_2737052
chicken anti-GFP	abcam	Cat# ab13970, RRID:AB_300798
Alexa Fluor 488 goat-anti-rat	abcam	Cat# ab150165, RRID:AB_2650997
Alexa Fluor 568 goat-anti-rat	abcam	Cat# ab175710, RRID:AB_2832918
Alexa Fluor 488 goat-anti-rabbit	abcam	Cat# ab150081, RRID:AB_2734747
CF405S goat-anti-mouse IgG1	Biotium	Cat # 20380
CF633 goat-anti-rabbit	Biotium	Cat# 20123-1, RRID:AB_10853138
CF633 goat-anti-mouse	Biotium	Cat# 20341
Alexa Fluor 488 goat-anti-mouse	Invitrogen	Cat# A-11029, RRID:AB_2534088
Alexa Fluor 568 goat-anti-mouse IgG1	Invitrogen	Cat# A-21124, RRID:AB_2535766
Alexa Fluor 488 goat-anti-mouse IgG2a	Invitrogen	Cat# A-21131, RRID:AB_2535771
Alexa Fluor 546 goat-anti-mouse IgG2a	Invitrogen	Cat# A-21133, RRID:AB_2535772
Alexa Fluor 568 goat-anti-guinea pig	Invitrogen	Cat# A-11075, RRID:AB_2534119
Alexa Fluor 488 goat-anti-rabbit	Invitrogen	Cat# A-11034, RRID:AB_2576217
Alexa Fluor 568 goat-anti-rabbit	Invitrogen	Cat# A-11036, RRID:AB_10563566
Alexa Fluor 647 goat-anti-rabbit	Invitrogen	Cat# A-21244, RRID:AB_2535812
Alexa Fluor Plus 488 goat-anti-chicken	Invitrogen	Cat# A32931, RRID:AB_2762843
Alexa Fluor 488 goat-anti-chicken	Invitrogen	Cat# A32931, RRID:AB_2762843
Atto 647N FluoTag X2 anti-mouse IgG1	NanoTag Biotechnologies	Cat# N2002-Atto647N
AbberiorStar635P FluoTag X4 anti-rabbit	NanoTag Biotechnologies	Cat# N2404-Ab635P
Alexa Fluor 488 goat-anti-rat	Jackson ImmunoResearch Labs	Cat# 112-545-167, RRID:AB_2338362
Alexa Fluor 488 goat-anti-chicken	Jackson ImmunoResearch Labs	Cat# 103-545-155, RRID:AB_2337390
Alexa Fluor 488 goat-anti-mouse	Jackson ImmunoResearch Labs	Cat# 115-545-166, RRID:AB_2338852
ChromoTek V5-Trap® Magnetic Agarose	ProteinTech	Cat# v5tma
ALFA Selector PE magnetic agarose affinity resin	NanoTag Biotechnologies	Cat# N1515
GFP Selector	NanoTag Biotechnologies	Cat# N0315
mouse anti-αTub	DSHB	Cat# 4A1
StarBright Blue 520 goat anti-mouse	Bio-Rad	Cat# 12005867
StarBright Blue 700 goat anti-rabbit	Bio-Rad	Cat# 12004162
Peroxidase-AffiniPure F(ab’)2 Fragment Goat Anti-Mouse	Jackson ImmunoResearch Labs	Cat#115-036-003, RRID:AB_2338518
AffiniPure Fab Fragment Goat Anti-Rabbit IgG, Fc Fragment Specific	Jackson ImmunoResearch Labs	Cat# 111-007-008, RRID: AB_2632459
Chemicals, Peptides, and Recombinant Proteins
Schneider’s Drosophila Medium	Gibco	Cat# 21720001
NEBuilder® HiFi DNA Assembly	NEB	Cat# E2621
Tris Base	Fisher Scientific	Cat# BP152-500
GAF fixative solution	Addax Biosciences	Cat# VI25
Triton™ X-100	Sigma-Aldrich	Cat# T8787
Goat Serum Donor Herd	Sigma-Aldrich	Cat# G6767
Glyoxal solution	Sigma-Aldrich	Cat# 128465
DPX mountant	EMS	Cat# 13510
Everbrite mounting medium	Biotium	Cat# 23001
Sodium azide	Sigma-Aldrich	Cat# S8032
Acryloyl-X, SE	Invitrogen	A20770
Anhydrous DMSO	Sigma-Aldrich	Cat # 276855
Poly-L-Lysine solution	Ted Pella	Cat# 18026
Photo-Flo detergent	EMS	Cat# 74257
Acrylic acid	Sigma-Aldrich	Cat# 147230
Acrylamide	Sigma-Aldrich	Cat# A9099
N,N’-Methylenebisacrylamide	Sigma-Aldrich	Cat# M7279
Sodium chloride solution	Sigma-Aldrich	Cat# S5150
PBS (10X), pH 7.4	Gibco	Cat# 70011044
4-Hydroxy-TEMPO	Sigma-Aldrich	Cat# 176141
N,N,N’,N’-Tetramethylethylenediamine	Sigma-Aldrich	Cat# T7024
Ammonium persulfate	Sigma-Aldrich	Cat# A3678
Proteinase K	NEB	Cat# P8107S
DAPI	Sigma-Aldrich	Cat# D9542
complete™, Mini Protease Inhibitor Cocktail	Roche	Cat# 11836153001
sodium deoxycholate	Sigma-Aldrich	Cat# BCCG2249
2x Laemmli Sample Buffer	Bio-Rad	Cat# 1610737
tris(2-carboxyethyl)phosphine	Goldbio	Cat# TCEP1
iodoacetamide	Sigma-Aldrich	Cat# I1149
Endoproteinase LysC	NEB	Cat# P8109S
Trypsin Protease, MS Grade	Thermo Scientific	Cat# 90057
ReproSil-Pur beads with 120 Å pores	Dr. Maisch	Cat# R119.AQ.0001
Acetonitrile, Optima™ LC/MS Grade	Fisher Chemical	Cat# A955
Immobilon Western Chemiluminescent HRP Substrate	Millipore	Cat# WBKLS
SeTau-647-NHS	SETA BioMedicals	Cat# K9-4149
Experimental Models: Organisms/Strains
Rdl Df: w[1118];; Df(3L)ED4421, P{w[+mW.Scer\FRT.hs3]=3’.RS5+3.3’}ED4421/TM6C, cu[1] Sb[1]	BDSC	RRID:BDSC_8066
GluClα Df: w[1118];; Df(3R)ED6025, P{w[+mW.Scer\FRT.hs3]=3’.RS5+3.3’}ED6025/TM6C, cu[1] Sb[1]	BDSC	RRID:BDSC_8964
Rdl mutant: Rdl[1]/TM3, Sb[1]	BDSC	RRID:BDSC_1687
w[1118]; P{y[+t7.7] w[+mC]=3XUAS-KDR.PEST}attP40	BDSC	RRID:BDSC_55791
y[1] w[67c23]; sna[Sco]/CyO, P{w[+mC]=Crew}DH1	BDSC	RRID:BDSC_1092
w[1118]; P{w[+mC]=GAL4::VP16-nanos.UTR}CG6325[MVD1]	BDSC	RRID:BDSC_4937
yw; sna[Sco]/CyO, P{w[+mC]=Crew}DH1	BDSC	RRID:BDSC_1092
w1118	BDSC	BDSC Cat# 3605; RRID:BDSC_3605
Mmd mutant: Slrp1[1]	BDSC	BDSC Cat# 42249; RRID:BDSC_42249
mmd-V5: w, mmd[Tag:V5.FRT]	BDSC	BDSC Cat# 95303; RRID: BDSC_95303
w;; GluClα-FlpStop.D	Ref.94	N/A
w;; nAChRβ1-KDRT-STOP-KDRT-smGdP-10xHA/TM6B	This study	N/A
w;; nAChRα1-KDRT-STOP-KDRT-smGdP-10xOllas/TM6B	This study	N/A
yw, nAChRα3-KDRT-STOP-KDRT-1xALFA;;	This study	N/A
w; nAChRalpha5-KDRT-STOP-KDRT-smGdP-10xOllas/CyO;	This study	N/A
w; nAChRalpha6-KDRT-STOP-KDRT-smGdP-10xOllas/CyO;	This study	N/A
w, nAChRα7-KDRT-STOP-KDRT-1xALFA;;	This study	N/A
w, nAChRα7-KDRT-STOP-KDRT-smGdP-10xOllas;;	This study	N/A
w;; Rdl-KDRT-STOP-KDRT-smGdP-10xV5/TM6B	This study	N/A
w;; Rdl-KDRT-STOP-KDRT-smGdP-10xHA/TM6B	This study	N/A
w;; Rdl-KDRT-STOP-KDRT-1xALFA/TM6B	This study	N/A
w, Lcch3-KDRT-STOP-KDRT-smGdP-10xOllas;;	This study	N/A
w;; Grd-KDRT-STOP-KDRT-1xALFA	This study	N/A
yw, CG8916-KDRT-STOP-KDRT-smGdP-10xV5;;	This study	N/A
w;; GluClα-KDRT-STOP-KDRT-smGdP-10xV5/TM6B	This study	N/A
w;; GluClα -KDRT-STOP-KDRT-1xALFA/TM6B	This study	N/A
w;; nAChRβ1-KDRT-smGdP-10xHA	This study	N/A
w;; nAChRα1-KDRT-smGdP-10xOllas	This study	N/A
yw, nAChRα3-KDRT-1xALFA;;	This study	N/A
w; nAChRalpha5-KDRT-smGdP-10xOllas;	This study	N/A
w; nAChRalpha6-KDRT-smGdP-10xOllas;	This study	N/A
w, nAChRα7-KDRT-1xALFA;;	This study	N/A
w, nAChRα7-KDRT-smGdP-10xOllas;;	This study	N/A
w;; Rdl-KDRT-smGdP-10xV5/TM6B	This study	N/A
w;; Rdl-KDRT-smGdP-10xHA/TM6B	This study	N/A
w;; Rdl-KDRT-1xALFA/TM6B	This study	N/A
w, Lcch3-KDRT-smGdP-10xOllas;;	This study	N/A
w;; Grd-KDRT-1xALFA	This study	N/A
yw, CG8916-KDRT-smGdP-10xV5;;	This study	N/A
w;; GluClα-KDRT-smGdP-10xV5/TM6B	This study	N/A
w;; GluClα -KDRT-1xALFA/TM6B	This study	N/A
w;; P{w+, 13xLexAoP-FRT-STOP-FRT-myrGFP-2A-KDR.PEST}attP1	This study	N/A
w, P{w+, 10xUAS-FRT-STOP-FRT-myrGFP-2A-KDR.PEST}attP40	This study	N/A
P{w+, 10xUAS-FRT-STOP-FRT-myrTdt-2A-KDR.PEST}attP5	This study	N/A
w, 10xUAS-myrTdtomato;;	BDSC	BDSC Cat# 32223; RRID: BDSC_32223
w, Cac-1xALFA;;	Bhukel et al., manuscript in preparation
w, P{w+, hsFLPG5.PEST}attP3;;	BDSC	BDSC Cat# 62118; RRID: BDSC_62118
w, P{w+, hs-FLPG5.PEST.Opt)attP3;;	BDSC	BDSC Cat# 77140; RRID: BDSC_77140
w;PBac{brp-RSRT-STOP-RSRT-smGdP-10xV5-2A-LexA-VP16}VK00001;	Ref.95
MBON-γ1pedc>α/β: sGal4MB112C: w;; P{w+, R13F04-GAL4.DBD}attP2, PBac{w+, R93D10-p65.AD}VK00027	BDSC	BDSC Cat# 68325; RRID: BDSC_68325
MBON-γ4>γ1γ2, MBON-β1>α: sGal4MB434B: w; P{w+, R30E08-p65.AD}attP40/CyO; P{w+, R53C10-GAL4.DBD}attP2	BDSC	BDSC Cat# 68325; RRID: BDSC_68325
Kenyon Cells sGal4MB010C: w; P{w+, R13F02-p65.AD}attP40; P{w+, R52H09-GAL4.DBD}attP2	BDSC	BDSC Cat# 68293; RRID: BDSC_68293
MBONα3, MBONα’2 sGal4MB082C: w;; P{w+, R23C06-GAL4.DBD}attP2, PBac{w+, R40B08-p65.AD}VK00027	BDSC	BDSC Cat# 68286; RRID: BDSC_68286
L5-gal4R64B07: w;; P{w+, GMR64B07-Gal4}attP2	BDSC	BDSC Cat# 39293; RRID: BDSC_39293
Tm3-gal4R13E12: w;; P{w+, GMR13E12-Gal4}attP2	BDSC	BDSC Cat# 48569; RRID: BDSC_48569
T4T5-gal4R42F06: w;; P{w+, GMR42F06-Gal4}attP2	BDSC	BDSC Cat# 41253; RRID: BDSC_41253
T4T5-LexAR42F06: w; P{w+, GMR42F06-LexA}attP40	BDSC	BDSC Cat# 54203; RRID: BDSC_54203
Mi9 sGal4SS02432: w; P{w+, R48A07-p65.AD}attP40; P{w+, VT046779-GAL4.DBD}attP2	BDSC	BDSC Cat# 86854; RRID: BDSC_86854
Recombinant DNA
pJFRC164-21XUAS-KDRT>-dSTOP-KDRT>-myr::RFP	Ref.70	RRID:Addgene_32141
pHD-DsRed-attP-w+	Gift from Kate O’Connor-Giles	RRID:Addgene_80898
pJFRC206-10XUAS-FRT>STOP>FRT-myr::smGFP-V5	Ref.67	RRID:Addgene_63168
pJFRC201-10XUAS-FRT>STOP>FRT-myr::smGFP-HA	Ref.67	RRID:Addgene_63166
pJFRC210-10XUAS-FRT>STOP>FRT-myr::smGFP-OLLAS	Ref.67	RRID:Addgene_63170
pJFRC161-20XUAS-IVS-KD::PEST	Ref.70	RRID:Addgene_32140
pJFRC177-10XUAS-FRT>-dSTOP-FRT>-myr::GFP	Ref.70	RRID:Addgene_32149
pJFRC19-13XLexAop2-IVS-myr::GFP	Ref.38	RRID:Addgene_26224
pJFRC5-5XUAS-IVS-mCD8::GFP	Ref.38	RRID:Addgene_26218
pBluescript II KS(-)	Agilent	Cat# 212208
pJFRC-5XUAS-GluClα-1XALFA	This study	N/A
pJFRC-5XUAS-Rdl-1XALFA	This study	N/A
pJFRC-10XUAS-FRT-STOP-FRT-myrGFP-2A-KDR::Pest	This study	N/A
pJFRC-10XUAS-FRT-STOP-FRT-myrTdt-2A-KDR::Pest	This study	N/A
pJFRC-13XlexAoP-10XUAS-FRT-STOP-FRT-myrGFP-2A-KDR::Pest	This study	N/A
pJFRC-13XlexAoP-10XUAS-FRT-STOP-FRT-myrTdt-2A-KDR::Pest	This study	N/A
Software and Algorithms
Zen Blue 2.3	Carl Zeiss	N/A
Zen Black 3.1	Carl Zeiss	N/A
neuTube	Ref.84	N/A
VVD Viewer	JaneliaSciComp	N/A
MaxQuant	Ref.92	N/A
FIJI ImageJ	Ref.88	N/A
Imaris 9.8	Oxford Instruments	N/A
R	Ref.86	N/A
python	Python software foundation	N/A
Adobe Illustrator	Adobe	N/A
Other
Nunc™ MiniTrays with Nunclon™ Delta surface	Thermo Scientific	Cat# 163118
Precision Cover Glasses, #1.5H Thickness	Thorlabs	Cat# CG15CH2
Bondic UV-curing adhesive	Bondic	Bondic starter kit
custom fabricated sample holder	Janelia Tech	ID 2021-021
Press-to-Seal™ Silicone Isolator with Adhesive, one well, 20 mm diameter, 0.5 mm deep	Invitrogen	Cat# P24740
Modified BP emission filter 527/23	Chroma	Cat# ET672/23m
Bel-Art mini-sieve	Bel-Art	Cat# F378451000
Porcelain Mortar	Fisher Scientific	FB961C
Porcelain pestle	Fisher Scientific	FB961M
Potter-Elvehjem tissue grinder	Cole Parmer	Cat # EW-04468-14
Steadystir digital S56	Fisher Scientific	N/A
Eppendorf™ Thermomixer™ R	Fisher Scientific	Cat# 05-400-205
Polymicro Flexible Fused Silica Capillary Tubing, Inner Diameter 75μm, Outer Diameter 375μm, TSP075375	Molex	Cat# 1068150019
SurePAGE™, Bis-Tris, 10x8, 4–20%, 15 wells	GenScript	Cat# M00657
Power Blotter Select Transfer Stacks, nitrocellulose, mini	Invitrogen	Cat# PB3210
Zeba™ Spin Desalting Columns, 7K MWCO, 0.5 mL	Thermo Scientific	Cat# 89882
TetraSpeck™ Microspheres, 0.2 μm, fluorescent blue/green/orange/dark red	Invitrogen	Cat# T7280

METHOD DETAILS

Identification of NR subunit genes

NR subunit gene numbers were obtained from https://www.genenames.org (human), https://www.informatics.jax.org (mouse) or https://www.flybase.org (fly).^62–64 For each organism, Cys-loop NR subunits, iGluR subunits, and GPCR neurotransmitter receptor subunits were compiled. GPCRs responding to neuropeptides were omitted from the counts.

Selection of tag insertion sites

All tags were inserted within the unstructured intracellular loop between the third (M3) and fourth (M4) transmembrane domains. In addition, we chose insertion sites that were poorly conserved and favored ones with evidence of sequence insertion in other species. We generated sequence alignments with the Clustal Omega program in UniProt (https://www.uniprot.org/align).⁶⁵ Closely related (e.g., D. yakuba) and distantly related (e.g., T. castaneum) insect species were used for alignment. The insertion sites are indicated in Figure S1 and summarized in Table S1. Species abbreviations - DROME: Drosophila melanogaster; DROSE: Drosophila sechellia; DROSI: Drosophila simulans; DROYA: Drosophila yakuba; DROAN: Drosophila ananassae; DROPS: Drosophila pseudoobscura; DROPE: Drosophila persimilis; DROWI: Drosophila willistoni; DROMO: Drosophila mojavensis; DROGR: Drosophila grimshawi; AEDAE: Aedes aegypti; ANOGA: Anopheles gambiae; CULSO: Culicoides sonorensis; TRICA: Tribolium castaneum.

Molecular biology

Generation of conditional tag cassettes

We generated a conditional tag cassette, pBS-KDRT-STOP-loxP-3XP3::dsRed-loxP-STOP-KDRT-smGFPTag, for the tags smGdP-10XV5, smGdP-10XHA, and smGdP-10XOllas. To generate the cassette, we used a combination of restriction enzyme-based cloning and HiFi DNA Assembly (NEB cat #E2621). First, we replaced the sequence between MluI and MfeI in KDRT-STOP-STOP-KDRT from pJFRC164⁷⁰ (Addgene plasmid #32141) with loxP-3XP3::DsRed-loxP from pHD-DsRed-attP-w+ (Addgene plasmid #80898). The 3xP3::DsRed marker allowed easy screening for successful genomic insertions, which are subsequently removed by germline expression of Cre recombinase. Next, we added a Drosophila codon-matched GS linker and the coding sequence for each tag downstream of KDRT. We cloned the sequence for each tag from the following plasmids: pJFRC206⁶⁷ (Addgene plasmid #63168) for smGdP-10XV5, pJFRC201⁶⁷ (Addgene plasmid #63166) for smGdP-10XHA, and pJFRC210⁶⁷ (Addgene plasmid #63170) for smGdP-10XOllas. To place the first KDRT sequence in frame, the cassette was preceded by the dinucleotide GG to encode a glycine with the first nucleotide of the KDRT sequence, and then the entire cassette was cloned into pBlueScriptII KS(−) (Agilent Cat# 212208) between PciI and XbaI. A second PciI site was inserted upstream of XbaI to facilitate cassette linearization in subsequent steps. All plasmids were sequence validated by Sanger sequencing. smGdPTag encodes a GFP protein with 10X epitope tags distributed amongst the C and N termini and one of the loops.⁶⁶ In addition to the inserted tags, the GFP sequence of smGdP contains amino acid substitutions that render GFP non-fluorescent (GdP: fluorescence dead GFP⁶⁷). For simplicity, we refer to smGdPTag throughout the study as smGFP-Tag or smTag.

We generated a conditional 1XALFA cassette by replacing smGdP-X from pBS-KDRT-STOP-loxP-3XP3::dsRed-loxP-STOP-KDRT-smGFP-Tag with Drosophila codon-optimized 1XALFA followed by a GS linker. ALFA-tag is a commercially developed epitope tag that forms a small and stable α-helix and is recognized by a high-affinity nanobody.⁶⁸ Detailed protocols are available upon request. The plasmids and sequences have been deposited in Addgene.

Generation of pU6-gRNA

We identified gRNA target sequences that cut within 1–11 nt of the selected insertion site with an efficiency score above 5, as defined by the CRISPR Efficiency Predictor (https://www.flyrnai.org/evaluateCrispr/). The gRNA sequence oligos were synthesized (Integrated DNA Technologies) with the forward oligo having a TTCG overhang at the 5’ end and the reverse oligo having an AAAC overhang added to the 5’ end for subsequent ligation into pU6. After annealing, the oligos were ligated into BbsI-linearized pU6b-sgRNA-short⁶⁹. All pU6 vectors generated were verified by Sanger sequencing. The gRNA sequences used in this study are listed in Table S1. gRNA1, used for generating alleles using short homology arms³¹ (see below), was cloned into pU6 as described for gene-specific gRNAs.

Generation of donor constructs

The generation of donor constructs involved two different methods for different sets of alleles. For the first set, long homology arms of approximately 1kb were used for homologous recombination of the conditional tag cassette. To improve cloning efficiency as well as donor integration by homologous recombination, we took advantage of a second method that employs shorter homology arms of around 100bp in combination with in vivo directed linearization of the donor vector, as described.³¹ This strategy was used to generate alleles of Grd, CG8916, and nAChRα3, as well as the 1XALFA tagged allele of GluClα.

For the generation of alleles using long homology arms, we employed HiFi DNA assembly (NEB Cat# E2621) to assemble the donor constructs. The long homology arms (~1kb) were PCR amplified and inserted into pHD-DsRed-attP-w+ (Addgene plasmid #80898), which was linearized with XhoI and EcoRI. The tag cassette was introduced by cloning in the PciI linearized conditional tag cassette of choice. For smGFP tags, we included a GS linker in the primer used to generate the 3’ homology arm. In contrast, the 1XALFA tag contains the GS linker within the conditional cassette. All pHD-geneX donor plasmids were sequence validated. Single nucleotide polymorphisms (SNPs) were permitted in intronic regions and in coding regions only when leading to synonymous codon substitutions from the dm6 reference genome.

For the generation of alleles using short homology arms, we utilized HiFi DNA assembly (NEB Cat# E2621) with two steps. In the first step, we synthesized the donor homology arms into pUC57-Kan (Genewiz, Inc.), with each of the two ~125bp homology arms flanked on the outside by gRNA1 target and PAM sequences and DNA assembly-specific homology arms matched to the chosen PciI linearized conditional cassette. The two arms were separated by a random sequence linker flanked by restriction sites that were absent in the homology arms. We included a GS linker in the synthesized sequence for the generation of donors with smGFP tags, while the 1XALFA conditional cassette included the GS linker within the cassette. In the second stage, we linearized pUC57-Kan-geneX with restriction enzymes recognizing the sites within the linker and cloned in the PciI linearized fragment that encoded the required conditional cassette using HiFi DNA assembly (NEB Cat# E2621). All donor vectors were confirmed by Sanger sequencing. Plasmids and sequences are available upon request.

Generation of pJFRC-10XUAS-FRT-STOP-FRT-myrFP-2A-KDR::Pest

HIFI DNA assembly (NEB Cat# E2621) was used to generate either 10XUAS-FRT-STOP-FRT-myr::GFP-2A-KDR::PEST or 10XUAS-FRT-STOP-FRT-myr::tdTomato-2A-KDR::PEST. The GFP coding sequence of pJFRC177⁷⁰ (Addgene: 10XUAS-FRT-STOP-FRT-myrGFP, plasmid #32149) was replaced either by GFP-2A (cassette C: GS linker-FRT-STOP-FRT-GFP-2A-LexAVP16⁷¹), or tdTomato-2A (UAS-DIPalpha-2A-tdTomato⁷²), both followed by the coding sequence of KDR::PEST recombinase from pJFRC161⁷⁰ (Addgene: 20XUAS-IVS-KD::PEST plasmid #32140). Plasmids are available in Addgene.

Generation of pJFRC-13XlexAoP-10XUAS-FRT-STOP-FRT-myrFP-2A-KDR::Pest

A combination of HIFI DNA assembly (NEB) and restriction enzyme-based cloning was used to generate 13XLexAoP2-FRT-STOP-FRT-myr::GFP-2A-KDR::PEST through modification of pJFRC177 ⁷⁰ (Addgene: 10XUAS-FRT-STOP-FRT-myrGFP, plasmid #32149). First, the 10XUAS sequence of pJFRC177 was replaced by 13XLexAoP2 from pJFRC19³⁸ (addgene:13XLexAoP2-IVS-myrGFP, plasmid #26224). Second, the GFP coding sequence of pJFRC177 was replaced either by GFP-2A (cassette C: GS linker-FRT-STOP-FRT-GFP-2A-LexAVP16⁷¹) followed by the coding sequence of KDR::PEST recombinase from pJFRC161⁷⁰ (Addgene: 20XUAS-IVS-KD::PEST plasmid #32140). Plasmids are available in Addgene.

Generation of pJFRC-5XUAS-GluClα-1XALFA and pJFRC-5XUAS-Rdl-1XALFA

We synthesized the coding sequence of GluClα-RM (protein isoform FBpp0307404 in FB2023_05) and Rdl-RA (protein isoform FBpp0076261 in FB2023_05). These sequences included a GS linker-1xALFA tag, which were placed at the same locations as in the respective conditionally tagged alleles (Table S1). The fragments were cloned into pJFRC5³⁸ (Addgene: 5XUAS-IVS-mCD8::GFP #26218), by replacement of the mCD8::GFP coding sequence. Synthesis and cloning were carried out by Genewiz, Inc.. Plasmids and sequences are available upon request. Flies were generated by injecting plasmid into embryos for recombination into attP5 sites by BestGene, Inc..

Generation of tagged NR subunit alleles

For alleles generated using donor vectors with long homology arms, pHD-geneX donor plasmid, and pU6-geneX-gRNA were injected into flies expressing Cas9 in the germline by BestGene Inc. Successful integration of the donor cassette was identified through the expression of DsRed in the eyes and negative for expression of mini-white in the eye. The resulting flies were PCR-validated for correct insertion within the selected insertion site (Table S1). To generate the conditional tagged allele, we crossed flies to a line expressing Cre recombinase (RRID:BDSC_1092) to excise loxP-flanked DsRed from the STOP cassette within the targeted allele (Figure S1C). To generate the whole fly constitutively tagged allele, we crossed the conditional tagged allele to a line expressing KD recombinase in the germline (w; 3XUAS-KDR; nos-gal4::VP16; RRID:BDSC_1092 and RRID:BDSC_4937) to excise the KDRT flanked STOP cassette upstream of the tag within the targeted allele. Resulting conditional and whole fly constitutively tagged alleles were balanced. When the whole fly constitutively tagged allele was not homozygous viable, we backcrossed flies three times with w1118 to clean out any off-target CRISPR events that could underlie lethality. We were not able to outcross homozygous lethality for Rdl-KDRT-smV5, Rdl-KDRT-smHA, and GluClα-KDRT-smV5. To test if lethality was due to the insertion of the tag within the NR subunit locus or to off-target CRISPR-induced mutations, we tested viability over deficiency lines (Rdl Df: RRID:BDSC_8066, GluClα Df: RRID:BDSC_8964) or loss of function alleles for Rdl (Rdl¹, RRID:BDSC_1687). As these animals were viable, lethality in homozygous animals does not reflect disruption of receptor function due to the tag. All tagged alleles were sequence validated to confirm the correct cassette excision. SNPs were allowed in intronic regions and in coding regions only when resulting in synonymous codon substitutions.

When generating alleles using short homology arms, pUC57-geneX donor plasmid, pU6-geneX-gRNA, and pU6-gRNA1 were injected into flies expressing Cas9 in the germline by BestGene Inc. Generation of flies was carried out as described above with the exception that flies were only screened for expression of DsRed in the eye. Conditionally tagged and whole fly tagged alleles of NR subunits were generated as described above.

Guidance for optimal sample processing and imaging of NR subunits in single neurons

The detection of endogenously tagged protein in single neurons requires optimal imaging and staining to achieve sufficient Signal-to-Noise Ratio (SNR) for the detection of receptor puncta in single neurons. In our experience, primary antibodies against V5 and HA provide the best SNR, making these epitopes preferable for tagging low-abundance proteins. Secondary antibodies often show significant lot-to-lot variability in SNR. Due to the lot-to-lot variation in polyclonal secondary antibodies, we recommend testing multiple antibodies to identify those with optimal SNR. We have summarized the antibodies used in this study and provided information on combinations that, in our experience, resulted in the best SNR (Table S4). It is also important to note that, as previously reported⁷³, immunostaining of many NR subunits is often unsuccessful using conventional 4% PFA fixation. To address this, we have developed a robust protocol employing acid-free glyoxal supplemented with sucrose (addaxS). This method enables robust detection of NR subunit puncta and ensures optimal preservation of neuronal morphology. We have also successfully used a previously published protocol for glyoxal fixation⁷⁴, which is comparable to addaxS for the detection of receptor puncta, but provides decreased preservation of fine neuronal processes, such as the dendrites of T4 neurons.

Optimal imaging conditions are crucial for achieving the best SNR and the resolution necessary to resolve receptor puncta while minimizing bleaching. Mounting samples in DPX is optimal for imaging with high numerical aperture (NA) objectives due to its matched refractive index and sample clearing properties, which are essential for detecting receptor puncta with the necessary resolution. In our experience, Airyscan played a crucial role in enhancing SNR and imaging resolution, enabling robust detection of NR subunits puncta with reduced sample bleaching. In contrast, conventional confocal microscopy often proved inadequate for visualizing these puncta, and when detected, they frequently underwent rapid bleaching. We provide details of mounting and imaging settings to achieve optimal Airyscan imaging, which are key to achieving robust imaging of NR puncta in single neurons (e.g. use of high-precision coverslips, DPX mounting, etc.). Additionally, ExLSM also improves SNR over conventional confocal microscopy due to the substantial reduction in background signal.”

Tagging of NR subunits in single neurons

Sparse labeling of cells was achieved as outlined in Figure S4B. To achieve sparse labeling of cells, we optimized the timing of heat shock to mediate Flp-out of the FRT-STOP-FRT (FSF) cassette from 10XUAS-FSF-myrFP-2A-KDR::Pest. To achieve sparse labeling of neurons found in each column or more than one per column, we used a less efficient variant of hsFlp (hsFlpG5::Pest(Opt); RRID:BDSC_77140). 0–24h APF pupae carrying hsFlp, 10XUAS-FSF-myrFP-2A-KDR::Pest, the cell type-specific GAL4 and the endogenous conditionally tagged allele of the NR subunit of interest (or more than one allele if more than one receptor was investigated, see Figures S5G–S5I), were heat-shocked at 37°C (Tm3:15 min, T4T5: 8–12 min, L5: 8 min) and subsequently reared at 25°C. The degree of labeled cells is very sensitive to small changes in time and temperature. The conditions for labeling were empirically established for every GAL4 driver. Brains for analysis were dissected from 1–5 day old flies. For developmental studies of NR subunit expression in T4/T5 and L5 neurons, the same procedure was followed as above, and brains were dissected at either 48h APF or 72h APF.

Expression of 5XUAS-GluClα-ALFA and 5XUAS-Rdl-ALFA in single T4 neurons (Figures S1A and S1B) was carried out using mosaic analysis with a repressible cell marker (MARCM).⁷⁵ 0–24h pupae were heat-shocked for 2 min at 37°C. ALFA-tagged constructs and myr::GFP were expressed in a restricted manner within single-cell clones under GAL4-UAS and were stained accordingly (see Table S3 for genotypes).

Immunostaining

For adult fly brains, 1–5 day old flies were decapitated, and the brain was dissected in ice cold Schneider Medium (SM) (Gibco, Cat# 21720001). Up to three brains were kept on ice in SM prior to fixation in a single well of a Terasaki plate (Thermo Scientific, Cat# 163118). All subsequent steps were carried out in Terasaki plates. Brains were fixed overnight at 4°C in glyoxal acid-free fixative (Addax Biosciences, Cat# VI25) supplemented with 5% (w/v) sucrose (addaxS) or in a fixative containing 3% glyoxal at pH 5.0 (3% glyoxal fixative)⁷⁴, as indicated in Table S3. When fixed with 3% glyoxal fixative, brains were fixed for 30 min at room temperature, followed by 30 min quenching in 0.1 M NH₄Cl. Fixation in acid-free glyoxal with 5% sucrose improves the preservation of fine neuronal morphology compared to 3% glyoxal fixative. Both fixatives improve immunostaining of NR subunits and Bruchpilot, compared to standard 4% PFA in PBS (some NR subunits, such as nAChRα7, cannot be visualized upon 4% PFA fixation). After fixation, brains were washed 3X and incubated for 2 hours at RT in PBSTX (PBS with 0.5% (v/v) Triton X-100 (Sigma Aldrich Cat# T8787) with the addition of 10% normal goat serum (PBSTN) (Sigma-Aldrich Cat# G6767). Brains were incubated for 2–3 days at 4°C with primary antibody mix in PBSTN, subsequently washed 2X 2hrs with PBSTX, and further incubated 2–3 days overnight with secondary antibodies in PBSTN. The secondary antibody was washed out by incubating brains 2X for 2 hrs in PBSTX and subsequently mounted as described below.

The following primary antibodies and concentrations were used for samples imaged using confocal microscopy. Rabbit anti-nAChRα6⁷³ (1:1,000), rat anti-nAChRα7⁷⁶ (1:1,000), rabbit anti-Rdl⁷⁷ (1:100), rabbit anti-GluClα (raised against a peptide from Musca domestica GluCl⁷⁸, which is conserved in D. melanogaster, 1:500), rabbit anti-Mmd (1:500 Guo et al., manuscript under preparation), rabbit anti-Ollas (1:10,000; GenScript, Cat# A01658), rat anti-Ollas (L2, Novus Biologicals Cat# NBP1-06713, RRID:AB_1625979), rat anti-HA (3F10, Roche Cat# 11867423001, RRID:AB_390918), mouse IgG2a anti-V5 (SV5-Pk1, 1:500; abcam Cat# ab27671, RRID:AB_471093), chicken anti-V5 (1:500, abcam Cat# ab9113, RRID:AB_307022), mouse anti-V5-Tag:DyLight^®550 (Bio-Rad Cat# MCA1360D550GA, RRID:AB_2687576), FluoTag X2 anti-ALFA::Atto488 (1:500; NanoTag Biotechnologies, Cat# N1502-At488), FluoTag X2 anti-ALFA::Alexa Fluor 568 (1:500; NanoTag Biotechnologies, Cat# N1502-AF568), FluoTag X4 anti-RFP::AZDye568 (1:500; NanoTag Biotechnologies, Cat# N0404-AF568), rabbit anti-DsRed (1:200, Takara Bio Cat# 632496, RRID:AB_10013483), guinea pig anti-RFP (1:1,000, Synaptic Systems, Cat# 390004, RRID:AB_2737052) and chicken anti-GFP (1:1000, abcam Cat# ab13970, RRID:AB_300798). Secondary antibodies were used at 1:500 for confocal microscopy. From abcam: Alexa Fluor 488 goat-anti-rat (Cat# ab150165, RRID:AB_2650997), Alexa Fluor 568 goat-anti-rat (Cat# ab175710, RRID:AB_2832918), and Alexa Fluor 488 goat-anti-rabbit (Cat# ab150081, RRID:AB_2734747). From Biotium: CF405S goat-anti-mouse IgG1 (Cat # 20380), CF633 goat-anti-rabbit (Cat# 20123-1, RRID:AB_10853138) and CF633 goat-anti-mouse (Cat# 20341). From Invitrogen: Alexa Fluor 488 goat-anti-mouse (Cat# A-11029, RRID:AB_2534088), Alexa Fluor 568 goat-anti-mouse IgG1 (Cat# A-21124, RRID:AB_2535766), Alexa Fluor 488 goat-anti-mouse IgG2a (Cat# A-21131, RRID:AB_2535771), Alexa Fluor 546 goat-anti-mouse IgG2a (Cat# A-21133, RRID:AB_2535772), Alexa Fluor 568 goat-anti-guinea pig (Cat# A-11075, RRID:AB_2534119), Alexa Fluor 488 goat-anti-rabbit (Cat# A-11034, RRID:AB_2576217), Alexa Fluor 568 goat-anti-rabbit (Cat# A-11036, RRID:AB_10563566), Alexa Fluor 647 goat-anti-rabbit (Cat# A-21244, RRID:AB_2535812), Alexa Fluor Plus 488 goat-anti-chicken (Cat# A32931, RRID:AB_2762843), and Alexa Fluor 488 goat-anti-chicken (Cat# A32931, RRID:AB_2762843). From NanoTag Biotechnologies: Atto 647N FluoTag X2 anti-mouse IgG1 (Cat# N2002-Atto647N) and AbberiorStar635P FluoTag X4 anti-rabbit (Cat# N2404-Ab635P). From Jackson ImmunoResearch Labs: Alexa Fluor 488 goat-anti-rat (Cat# 112-545-167, RRID:AB_2338362), Alexa Fluor 488 goat-anti-chicken (Cat# 103-545-155, RRID:AB_2337390) and Alexa Fluor 488 goat-anti-mouse (Cat# 115-545-166, RRID:AB_2338852).

Tagging of NR subunits in neurons of the mushroom body

The genotype used for tagging NR subunits in single neurons of the mushroom body is reported in Table S3. A 2 hr heat shock at 24–48 hrs APF was chosen to generate single neuron labeling. Dissection and immunohistochemistry of fly brains were carried out as previously described with 3% Glyoxal fixative instead of 2% PFA³⁷ using the antibodies listed in Table S3. Brains of 5-day-old female flies were dissected in Schneider’s insect medium and fixed in 3% Glyoxal fixative (3% v/v glyoxal (Sigma-Aldrich, Cat# 128465), 5% v/v ethanol, 0.75% v/v acetic acid, pH 5.0) for 2 hr at room temperature (RT). The samples were quenched in 0.1M NH₄Cl for 30 min, followed by four washes in PBT (0.5% Triton X-100 in 1x-PBS), 10 min each. After washing in PBT, tissues were blocked in 5% normal goat serum (or normal donkey serum, depending on the secondary antibody) for 90 min. Subsequently, tissues were incubated in primary antibodies diluted in 5% serum in PBT for 2 days on a nutator at 4°C, washed four times in PBT for 30 min, then incubated in secondary antibodies diluted in 5% serum in PBT for 2 days on a nutator at 4°C. Tissues were washed thoroughly in PBT four times for 30 min or longer.

Sample mounting for confocal microscopy

Samples were mounted using either DPX mountant (EMS, Cat# 13510) as previously described⁶⁷ or Everbrite mounting medium (Biotium, Cat# 23001), as denoted in Table S3. DPX facilitates the imaging of neurons deep in the brain. The refractive index of DPX matches the refractive index of the immersion oil used with high NA objectives, facilitating higher resolution achievable using Airyscan.

Confocal microscopy

All confocal images were acquired on a Zeiss LSM880 with 405 nm, 488 nm, 561 nm, and 633 nm lasers. Images were acquired with a Plan-Apochromat 63x/1.4 Oil DIC M27 objective for single cell and whole brain imaging or a Plan-Apochromat 40xx/1.2 Imm Korr DIC M27 for whole optic lobe imaging (Figure S1).

Airyscan was used in conjunction with 63X objective imaging in RS mode to maximize signal capture and increase resolution compared to conventional confocal imaging. All samples were imaged using precision cover glasses with #1.5H thickness (Thorlabs, Cat# CG15CH2), and the Airyscan detector alignment was verified and adjusted prior to imaging each sample. Frame size and Z step were optimal to achieve maximal resolution. Images were processed using Zen Blue 2.3 with Airyscan processing set to auto-filter and 3D processing.

ExLLSM sample preparation and imaging

The samples for Expansion Microscopy (ExM) were dissected, fixed, and immunostained as stated for neurons of the mushroom body, and staining was done sequentially to avoid cross-reactivity. All samples were processed using a protein retention ExM protocol with minor modifications.^79,80 All solutions were prepared in milliQ-grade water unless otherwise stated: AcX stock (acryloyl-X, SE Invitrogen, A20770) at 10 mg/mL in anhydrous DMSO (Sigma-Aldrich, Cat # 276855); PLL solution (Ted Pella, Cat# 18026) with Photo-Flo detergent (EMS, Cat# 74257) added 1:500 (v/v); Acrylate stock at 4M, prepared by neutralizing 5.5 mL acrylic acid (99% purity; Sigma-Aldrich, Cat# 147230) with 10N NaOH using a water bath and fume hood, in a total volume of 20mL⁸¹; Acrylamide stock at 50% (w/v) (Sigma-Aldrich, Cat# A9099); and Bis-acrylamide stock at 1% (w/v) (Sigma-Aldrich, Cat# M7279). Monomer stock: 11.5 mL sodium acrylate stock, 2.5 mL acrylamide stock, 7.5 mL bis-acrylamide stock, 18 mL 5 M NaCl (Sigma-Aldrich, Cat# S5150), 5 mL 10xPBS (Gibco, Cat# 70011044), and 2.5 mL water for a total volume of 47 mL. 4HT stock: 4-hydroxy-TEMPO at 0.5% (w/v) (Sigma-Aldrich, Cat# 176141). TEMED stock: N,N,N′,N′-Tetramethylethylenediamine at 10% (v/v) (Sigma-Aldrich, Cat# T7024). APS stock: ammonium persulfate at 10% (w/v) (Sigma-Aldrich, Cat# A3678). ProK digestion buffer: 0.5% Triton X-100, 500 mM NaCl, 1 mM EDTA, and 50 mM Tris pH8. Appropriate caution was exercised when handling acrylamide, a known toxin.

Dissected, fixed, and immunostained samples were treated with AcX stock solution diluted 1:100 in 1xPBS overnight. Brains were then washed with 1xPBS. A gelation chamber was created by applying a Press-to-Seal silicone gasket (Invitrogen, Cat# P24740) to a glass slide, which was then coated with the PLL solution. AcX-treated brains were immobilized on the PLL surface, up to nine per gasket. Gelation solution was prepared just before gelation to prevent premature gel polymerization on ice by adding 10 μL each of 4HT, TEMED, and APS stock solutions to 470 μL of monomer stock solution. Brains were washed with gelation solution, and then the gelation chamber was filled with ~200 μL of gelation solution and incubated on ice for 25 min. The gelation chamber was then sealed with a cover slip and placed in a 37°C incubator to gel and cure for 2 hr.

Gelation chambers were disassembled, and individual gels were trimmed close to each brain. Gels were trimmed to a right trapezoid shape to ease specimen orientation. Gels were incubated with proteinase K (NEB, Cat# P8107S) diluted 1:100 (v/v) in proK digestion buffer with shaking overnight at room temperature. Digested gels were stained in 500 ng/mL DAPI in PBS for 30 min, followed by four washes with water for 30min each, followed by equilibration overnight. Prepared ExM samples were stored in 1xPBS at 4°C for a week. The samples were expanded in MilliQ water for 3 hours for optimum expansion before imaging by LLSM. Samples expanded ~4.65x. All samples were scanned within 3–5 hr of expansion.

All ExM samples were imaged in objective scan mode as described²³ with minor modifications. For all imaging sessions, focus was maintained by periodic imaging of reference beads. The region of interest was identified by scanning the Brp channel with minimal exposure and acquired as a single tile of 1024×1024×501 voxels.

ExLSM sample preparation and imaging

Tissue staining and expansion for ExLSM

Tissue for ExLSM was expanded as described above with the following modifications. Fixation and staining of samples for ExLSM were performed with increased concentration of primary and secondary antibodies compared to samples prepared for confocal microscopy (see Table S3). Brains were incubated in gelation solution for 30 minutes at 4°C prior to transfer to the gelation chamber. Digested samples were washed in 1X PBS and stained with 1:1000 DAPI (Sigma-Aldrich, Cat# D9542) in 1X PBS for 30 minutes at room temperature unless antibodies with 405-dye conjugate were used. Samples were expanded to ~4X in autoclaved Milli-Q water at room temperature before mounting onto PLL-coated coverslips that were subsequently bonded with Bondic UV-curing adhesive (Bondic starter kit, Bondic) onto a custom fabricated sample holder (Janelia Tech ID 2021–021) to be suspended vertically in the imaging chamber. Mounted samples were imaged in 1mM Tris Base (Fisher Scientific, Cat# BP152-500) in MilliQ water after a minimum of 2 hours of incubation at room temperature or overnight at 4°C. Unexpanded gels were stored at 4°C in 1X PBS + 0.02% sodium azide (Sigma-Aldrich, Cat# S8032) for up to 10 days before expansion and imaging.

Light-sheet imaging on Zeiss LS7

Images were acquired on a Zeiss LS7 microscope equipped with 405 nm, 488 nm, 561 nm, and 638 nm lasers. Illumination optics with a 10x/0.2 NA were used for excitation (Zeiss, Cat# 400900-9020-000). Detection was performed using a W Plan-Apochromat 20x/1.0 DIC M27 water immersion objective (Zeiss, Cat# 421452-9700-000). The LS7 optical zoom was set to 2.5x, resulting in a total magnification of 50x. CF405S and AF546 dyes were simultaneously excited by the 405 nm and 561 nm laser lines, and emission light was separated by a dichroic mirror SBS LP 510 with emission filters BP 420–470 (Zeiss, Cat# 404900-9312-000) and a modified BP 527/23 (Chroma, Cat# ET672/23m). Similarly, AF488 and SeTau647 dyes were simultaneously excited via 488 nm and 638 nm, and the emission was split through a dichroic SBS LP 560 with emission filters BP 505–545 and LP 660 (Zeiss, Cat# 404900-9318-000). As the final pair, AF568 and SeTau647 dyes were excited together via the 561 nm and 638 nm laser lines, and emission was filtered by a dichroic SBS LP 640 with emission filters BP 575–615 and LP 660 (Zeiss, Cat# 404900-9322-000). To eliminate laser transmission, a 405/488/561/640 laser blocking filter (Zeiss, Cat# 404900-9101-000) was added to the emission path. Images were captured using dual PCO.edge 4.2 detection modules (Zeiss, Cat# 400100-9060-000) with a 50 msec exposure time. Filter and camera alignment were manually calibrated prior to each imaging session. Image volumes were acquired at optimal Z-step and light-sheet thickness, and the Pivot Scan feature was used to reduce illumination artifacts by sweeping the light-sheet in the xy-plane. The LS7 microscope was operated using ZEN Black 3.1 (v9.3.6.393).

Affinity purification and mass spectrometry

To define GluClα interactors, we used affinity purification and mass spectrometry (Figure 5A). To increase confidence of true interactor identification, we used two different tagged alleles of GluClα (GluClα-smV5 or GluClα-ALFA) and defined putative interactors as the factors that were identified in mass spectrometry of the immunoprecipitation (IP) with matched nanobody to that of the tag (ALFA or V5) (V5, ProteinTech, Cat# v5tma, ALFA selector, NanoTag Biotechnologies, Cat# N1515). As a negative control, we used the same nanobodies against the unmatched tagged allele (e.g., ALFA nanobody with GluClα-smV5 samples). Each set of conditions was done in biological triplicate (Figure 5A). We used stringent criteria to identify GluClα interactors where a putative interactor had to be significantly enriched in both experimental conditions over the negative controls (Figure 5A). In this way, we identified the GluClα interactor Mmd (Table S2).

Immunoprecipitation was performed on 2–10 day old flies of either GluClα-smV5 (w;; GluClα-smV5/+) or GluClα-ALFA (w;; GluClα-1XALFA/+) genotype, which were previously frozen in liquid nitrogen and stored at −80°C. For each pulldown, 10 ml of frozen flies were sieved using a Bel-Art mini-sieve (Bel-Art, Cat# F378451000). Heads were pulverized in liquid nitrogen using a liquid nitrogen chilled mortar and pestle (Fisher Scientific, Cat# FB961C & FB961M) and further homogenized in 1.5 ml of lysis buffer (50 mM Tris HCl (pH 7.6), 150 mM NaCl, 10% glycerol, protease inhibitor (cOmplete mini, Roche, Cat# 11836153001) using a Potter-Elvehjem tissue grinder (Cole Parmer, Cat # EW-04468–14) with 10 strokes at 900 rpm on a Steadystir digital S56 (Fisher Scientific) in a 4°C room on ice, spaced 10s apart to avoid sample overheating. Homogenate was centrifuged at 1,000 ×g for 10 min at 4°C to remove the soluble fraction. The insoluble fraction was resuspended in an equal volume of lysis buffer with 0.5% sodium deoxycholate (DOC, Sigma-Aldrich, Cat# BCCG2249) and rotated head-over-tail for 30 min at 4°C. The solubilized homogenate was centrifuged for 15 min at 16,000 ×g at 4°C. The supernatant was transferred to either V5-Trap magnetic agarose or ALFA selector PE agarose magnetic beads and washed 2 times in 1 mL of lysis buffer according to the manufacturer protocol. The sample and beads were incubated with head-over-tail rotation for 1 hour at 4°C. Samples were washed with lysis buffer containing 0.5% DOC according to the manufacturer’s instructions. Bound protein was eluted in 100 μl of 2X Laemmli buffer (Bio-Rad, Cat# 1610737) for 5 min at 95°C with 1,000 rpm shaking in a Thermomixer (Fisher Scientific, Cat# 05-400-205).

For proteomic characterization of affinity-purified tagged GluClα samples, eluates in 1X Laemmli buffer were diluted two-fold to reduce the SDS concentration. Samples were reduced and alkylated by the sequential addition of 5 mM tris(2-carboxyethyl)phosphine (Goldbio, Cat# TCEP1) and 10 mM iodoacetamide (Sigma-Aldrich, Cat# I1149). This was followed by protein clean-up using the single-pot, solid-phase-enhanced sample preparation (SP3) protocol.⁸² Subsequently, the samples underwent proteolytic digestion with Lys-C (NEB, Cat# P8109S) and trypsin (Thermo Scientific, Cat# 90057) at 37°C overnight. The resulting peptides were desalted using SP3-based peptide clean-up and analyzed by LC-MS/MS. Briefly, peptides were separated by reversed-phase chromatography using 75 μm inner diameter fritted fused silica capillary column (Molex, Cat# 1068150019) packed in-house to a length of 25 cm with bulk 1.9mM ReproSil-Pur beads with 120 Å pores (Dr. Maisch, Cat# R119.AQ.0001).⁸³ The increasing gradient of acetonitrile (Fisher Chemical, Cat# A955) was delivered by a Dionex Ultimate 3000 (Thermo Fisher Scientific) at a flow rate of 200nL/min. The MS/MS spectra were collected using data-dependent acquisition on Orbitrap Fusion Lumos Tribrid mass spectrometer (Thermo Fisher Scientific) with an MS1 resolution (r) of 120,000 followed by sequential MS2 scans at a resolution (r) of 15,000.

Co-immunoprecipitation and western blotting

The interaction between GluClα and Mmd was validated by repeating the above immunoprecipitation experiment using magnetic agarose beads with a nanobody recognizing GFP (GFP Selector, NanoTag, Cat# N0315), which interacts with protein harboring the V5 tag (smV5). We used w1118 and w; Rdl-smV5/+ flies as controls to assess the specificity of the interaction. IP was carried out as described above on 5 ml of flies, and the solubilized homogenate (input) and the eluted protein were run on a denaturing SDS-PAGE using SurePage Bis-Tris 4–20% gels (GenScript, Cat# M00657). Proteins were transferred onto a nitrocellulose membrane with semi-dry transfer (Invitrogen, Cat# PB3210), and membrane was blocked in TBST (10mM Tris-HCl, 0.9% w/v NaCl, 0.02% v/v Tween 20) with 5% skim-milk. Membranes were probed with mouse anti-V5 (SV5-Pk1, 1:500; Abcam Cat# ab27671, RRID:AB_471093), rabbit anti-Mmd (Guo et al., manuscript under preparation), and mouse anti-αTub (DSHB, Cat# 4A1). Anti-V5 was detected with StarBright Blue 520 goat anti-mouse (Bio-Rad, Cat# 12005867), anti-Mmd with StarBright Blue 700 goat anti-rabbit (Bio-Rad, Cat# 12004162), and anti-αTub with HRP conjugated goat-anti-mouse (Jackson ImmunoResearch Labs, Cat#115-036-003). HRP-conjugated antibodies were detected with a commercial chemiluminescence system (Millipore, Cat# WBKLS), while fluorescent dye-conjugated antibodies were directly detected on an iBright 1500 imaging system (Invitrogen).

Secondary antibody labeling

In-house dye conjugation of Fab goat anti-rabbit IgG Fc SeTau647 antibody was performed by cross-linking Fab fragment goat anti-rabbit IgG Fc (Jackson ImmunoResearch, Cat# 111-007-008, RRID: AB_2632459) with amine-reactive SeTau-647-NHS (SETA BioMedicals, Cat# K9-4149) at a 1:10 ratio with addition of 0.1M sodium bicarbonate. The mixture was incubated at room temperature with shaking at 500 rpm for 1 hour. Labeled antibody was passed through a Zeba spin desalting column (Thermo Scientific, Cat# 89882). Desalinated labeled antibody was diluted in 40% glycerol at a concentration of 1–5mg/mL and stored at 4°C.

QUANTIFICATION AND STATISTICAL ANALYSIS

EM connectome analysis and visualization

Coordinates of synapses and neuron skeletons for the Fib25 EM dataset were obtained online (https://github.com/janelia-flyem/ConnectomeHackathon2015). Only synapses with a confidence level of 1 were considered in the analysis. For Tm3 neurons, only fully reconstructed neurons (Tm3-home-ant, Tm3-B-ant (rep), and Tm3-C-ant) that fully arborized within the volume in which synapses were annotated were used. For L5 neurons, six L5 neurons (A-F) located in the columns surrounding the central home column, as well as the L5 neuron in the home column, were analyzed. The distribution of synapses from inputs that made three or more synapses onto L5 dendrites in the medulla was computed in each layer. Layer boundaries were determined using L1 neurons as reference (these neurons arborize in M1 and M5 layers). The neurotransmitter identity of each neuron input was obtained from scSeq data.⁸ The visualization of skeletons and synapses was done using neuTube.⁸⁴

To visualize T4 and T5 EM data and synaptic inputs, the coordinates of synapses and neuron skeletons for the Fib19 dataset⁴ were obtained by downloading the data using neuprintr⁸⁵. This was done for five representative reconstructions of T4 and T5 neurons.⁴ Specifically, for T4 neurons, inputs to the medulla were considered, while for T5 neurons, inputs to the lobula dendrites were analyzed. For Figure 7, all 20 reference T4 dendrites were visualized alongside the Mi1 and Mi9 Home and the ones from columns A-F. Synapses between the home column Mi1 and Mi9 and T4 dendrites were identified. Partners are shown if they made 5 or more synapses, except for a second T4c neuron, which was omitted in Figure 7C for simplicity but formed a similar number of synapses with Mi1 home in its middle domain as seen with T4c-2. The visualization of skeletons and synapses was done using neuTube.⁸⁴

Confocal image analysis

Airyscan processed images of L5, Tm3, T4, and T5 neurons were analyzed using Imaris 9.8. First, the membrane channel was extracted using the surface function to create a mask of the neuron. The receptor signal within the mask was analyzed using the spot function to identify puncta above background. All single neuron images shown are 3D projections of masked membrane and receptor signal. To determine the distribution of receptor puncta in T4/T5 dendrites, we restricted the analysis to c/d type T4 or T5 neurons where the direction of dendrite projection was on the long axis of the dendrite, as this facilitated the identification of the direction of dendritic processes. The proximo-distal axis of dendrites was normalized to 1, and the position of detected puncta was computed onto the normalized axis. All analyses were performed using R Statistical Software⁸⁶ (v4.2.2) using the ggplot2⁸⁷ package. All other images were processed using FIJI ImageJ.⁸⁸

A Wilcoxon rank sum test with continuity correction was used to obtain significance values when comparing T4, T5, L5, and Tm3 comparison to synapse counts in EM data. Bonferroni correction was applied to p values when comparing L5 and Tm3 synapse counts to EM data in the three distinct dendritic compartments.

Images were assembled into figures using Adobe Illustrator.

Quantification of NR subunit expression

For quantification of NR subunit expression in Figure S1G, for each brain, we acquired 5 frames of a central portion of the medulla spanning all ten layers in each optic lobe. We used FIJI imageJ to obtain the signal levels and computed the mean of all 5 frames and both optic lobes for each brain. The ratio of signal using an antibody against the native protein normalized with the signal of the pan synaptic Brp is shown. Different antibody dilutions were used to rule out saturation of the antibody signal at the standard antibody concentration typically used.

ExLLSM image analysis

Analysis of ExLLSM images was carried out as previously described²³ with minor modifications. All the datasets acquired were deconvolved with the Richardson-Lucy algorithm using experimentally measured point spread functions (PSFs) for each color channel for ten iterations²⁴. Only the central 360×480×480 voxels where LLSM provides optimal resolution were analyzed. After deconvolution, images were filtered to remove small puncta below 9 voxels by 3D Gaussian kernel with a sigma value of 1 and intensity threshold defined as the larger value between the mean plus two standard deviations and the value calculated with Otsu’s method. We applied a lower size threshold (9 instead of 1008 voxels) to include putative non-synaptic receptor signal that may appear as small puncta for analysis. The 3D local maxima above the intensity threshold were detected using 3D spheric kernels of 2 voxel-radius by running 3DIJ2⁸⁹. The images were segmented using the detected local maxima as seeds to run 3D watershed. The descriptive parameters of segmented objects, such as the center of mass (COM) and number of voxels, were measured using 3D ImageJ Suite⁹⁰. We defined colocalization between receptor and Brp objects if each segmented object of the receptor channel had at least one voxel that was within 1 voxel distance from Brp objects irrespective of the distance of the COMs. The Brp-colocalized receptor objects defined by this criterion appeared to be larger and closer to Brp compared to the “non-colocalized” receptor objects (Figures S3M–S3P). The scale shown for ExLLSM and ExLSM images is based on the expansion rate of 4.65, which was estimated by epifluorescence imaging of 16 brains before and after expansion. The expansion rate was larger than previously reported²³ when using PFA as fixative instead of Glyoxal-based fixatives.

ExLSM image analysis

ExLSM images were processed using Python (v3.9.13). Richardson-Lucy deconvolution was performed for 80 iterations on puncta channels stained for nc82 or smV5 and for 10 iterations on membrane channels stained for myr::GFP or myr::tdTomato. Experimental PSFs for deconvolution were acquired by imaging 200 nm fluorescent microspheres (Invitrogen, Cat# T7280), and an average PSF was extracted using the Experimental PSF Wizard in ZEN Blue (v3.4.91.00000). Deconvolved data was intensity normalized in the range of −1 to +40 standard deviations, and binary masks were generated by applying a Laplacian of Gaussian function and setting a positive threshold (0.01 for all myr::GFP and myr::tdTomato channels, 0.02 for all nc82 channels, and 0.1 for all smV5 channels). Puncta channels were then processed as follows. First, local intensity peaks were identified using a 3-voxel diameter window. To group multiple peaks located at the same synapse, nearest neighbors below a defined distance threshold were iteratively averaged, using an 8-voxel radius for Brp-smV5 or nc82 channels and a 4-voxel radius for GluClα-smV5. The processed peaks were used to perform a marker-controlled watershed transform on the masked image⁹¹, giving the final segmentation results.

Neuron masks were created through manual proofreading and labeling of the membrane channel masks in VVD Viewer (v1.5.10) (for selecting the desired neuron segments) and Napari (v0.4.17) (for manually painting the neuron labels). Nc82 signal within Mi9 axon terminals was extracted by calculating the uniformly weighted centers of mass (i.e. centroids) of each segmented object in the nc82 channel and selecting the objects whose centroids fell within the neuron mask.

All image panels displayed in Figure 6 and Figure S7 show 3D representations produced in VVD Viewer from analyzed data. Each panel shows a representative T4 dendrite from three different brain samples.

Analysis of mass spectrometry data

The data generated by LC-MS/MS were analyzed using the MaxQuant bioinformatic pipeline⁹². The Andromeda integrated in MaxQuant was employed as the peptide search engine, and the data were searched against the Drosophila melanogaster database (UniProt Reference UP000000803). A maximum of two missed cleavages was allowed. The maximum false discovery rate for peptide and protein was specified as 0.01. Label-free quantification (LFQ) was enabled with LFQ minimum ratio count of 1. The parent and peptide ion search tolerances were set to 20 and 4.5 ppm, respectively. The MaxQuant output files were subsequently processed for statistical analysis of differentially enriched proteins using Analytical R tools for mass spectrometry (artMS)⁹³.

Sc-RNAseq expression

Heatmaps of the expression of genes at 96h APF (Figure S4A) or during pupal development (Figure S6A) were obtained from previously published scSeq RNAseq data⁸ (GSE156455).

Statistical analyses

Identification of NR subunit puncta was done using Imaris 9.8 as outlined in the “Confocal image analysis” section of the STAR Methods. Statistical analysis was done using R Statistical Software⁸⁶, and details, such as the test used, values of n, and confidence intervals, are reported in the figure legend. For quantification of puncta in single neurons, n represents the number of individual neurons from at least 3 animals. For analysis of synapses in single EM datasets (see EM connectome analysis and visualization section of the STAR Methods), n represents individual neurons from a single brain. Throughout the study, boxplots represent the median (central line in the box) and interquartile range (box). For analysis of ExLLSM images (Figures 3M–P), 8–24 tiles were analyzed for each condition as described in the “ExLLSM Image Analysis” section of the STAR Methods from at least 2 animals. ExLLSM analysis was performed using Prism Software and the details of the analysis are reported in the Figure legend.

Supplementary Material

3

Table S3. Table of genotypes and staining conditions for each figure panel – Related to STAR Methods

Highlights.

• Conditional epitope-tagging of 11 neurotransmitter receptor (NR) subunits.

• NR subunits localize to specific domains within single dendrites.

• Specific NR subunits localize to cell-type specific synaptic connections.

• The membrane protein Mmd selectively associates with a glutamate receptor in vivo