A genetic strategy for targeting local astrocytes in adult Drosophila

Summary

Astrocytes are the brain’s major glial population and have been associated with a vast number of functions. To probe this diversity and to reach a similar level of understanding about astrocyte physiology that we have about neurons, we need genetic tools targeting specific astrocyte subpopulations. In Drosophila, available tools have historically been restricted to driver lines driving expression in astrocytes throughout the brain. To target specific astrocytes, we have optimized the genetic tool transneuronal control of transcription (TRACT; referred to as astro-TRACT), allowing effector expression specifically in local astrocytes of a given neuronal circuit. We analyzed the tool’s specificity, sensitivity, and reproducibility across various mushroom body (MB) split-Gal4 drivers. We found that the number of pre-synapses correlates positively with labeling efficiency and that single local astrocytes around MB medial compartments project into the ellipsoid body (EB). Astro-TRACT will be valuable to investigate mechanistic astrocyte-neuron signaling, functional, and structural astrocytic diversity across the adult fly brain.

Graphical abstract

Highlights

• •Optimized astro-TRACT targets local astrocytes in adult Drosophila
• •Labeling correlates with presynaptic density of the sending circuit
• •Astrocytes bridge the mushroom body and ellipsoid body

Motivation

The extent and refinement of available tools for studying glial cells remain limited compared to those developed for neurons. To advance our understanding of glia and their signaling mechanisms in interactions with neurons and other glia, methods that enable precise visualization and manipulation of targeted glia are necessary. Here, we have optimized a genetic tool in adult Drosophila to target subsets of the largest glial class, astrocytes, near user-defined neuronal circuits.

Dopp et al. optimized a genetic tool to target local astrocytes in adult Drosophila. They find that labeling efficiency depends on synaptic density. Using this tool, they show that astrocytes bridge the mushroom body and ellipsoid body, two brain centers that lack direct neuronal connections.

Introduction

Genetic tools have advanced our understanding of the brain on both structural and functional levels. They facilitate the control of specific cellular subpopulations and are necessary to address mechanistic neuroscientific questions across model systems.^1,2 Together with neurons, glial cells populate the brain, and astrocytes make up their largest class. In the past, astrocytes have been viewed as housekeeping cells. In contrast, astrocytes are active participants in many physiological processes such as synaptic plasticity,³ gliotransmission, cytokine signaling, recycling and buffering of neurotransmitters and ions⁴ and downstream behaviors such as sleep homeostasis.⁵ Specific astrocytes involved in regulating a given process may be located close to a relevant neuronal circuit. Understanding how such local astrocytes regulate processes requires functional studies.

In Drosophila, an extensive library of thousands of driver lines has been generated that facilitates precise genetic control of specific neuronal subpopulations.^6,7 Unfortunately, the creation of a similar library targeting glial subtypes is challenging. In contrast to mammalian astrocytes, transcriptionally there are little differences between fly astrocytes located in different regions of the brain.⁸ Furthermore, region-specific astrocytic drivers in the fly central brain have not been reported in the literature so far. Despite the lack of transcriptional diversity, the morphological diversity of astrocytes across the fly brain is well-documented.^8,9 For example, astrocytes around the mushroom body (MB) have unusual thin branches.⁹ The region-specific glial structural patterns have been attributed to specialized, dynamic subtypes that adapt in response to cues from their local environment,^8,9 and may serve different functions in local circuits.

Here, we describe another strategy for targeting local astrocyte subpopulations, which takes advantage of the extensive library of neuronal drivers instead of relying on genetic differences between astrocyte populations. We adapted the TRACT (transneuronal control of transcription) system, which is based on synthetic Notch-mediated transmembrane proteolysis.^10,11 It was initially developed to study neuronal wiring and cell-cell interactions, including glia-neuron contacts in the developing fruit fly. Huang et al. (2016) characterized glial expression across multiple larval brain regions and compared the expression between alrm and repo enhancers. Here, we have validated and optimized the TRACT tool for adult flies, analyzed its specificity, sensitivity and reproducibility and applied it to investigate morphological patterns of MB–associated astrocytes. We found that astrocytic labeling is achieved only when sufficient synapses are present in their associated MB compartment. Additionally, we found that local astrocytes of MB γ, β, and β′ compartments connect to the EB.

Results

Flexibility of astro-TRACT is increased by using split-Gal4 system to express ligand in specific sending neuronal circuits

The original TRACT system comprises four transgenes distributed across two binary expression systems: the LexA>LexAop system drives the expression of the ligand (CD19) for the synthetic Notch receptor in designated sending neurons, and the Gal4>UAS system is utilized in the synthetic Notch proteolysis system within the receiving cells. Therefore, the specificity of TRACT largely relies on the choice of driver used for ligand expression. However, the existing LexA-driver resource is limited in the number of available lines and their neuronal targeting specificity. In contrast, the split-Gal4>UAS system achieves significantly higher specificity. Split-Gal4 drivers consist of two independent components—VP16AD and Gal4DBD—that form a functional Gal4 protein only when co-expressed in cells containing both enhancers.⁷ A comprehensive library of split-Gal4 drivers targeting highly specific neuronal populations has been generated,^6,7 and additional unique neuronal populations can be readily targeted by combining different VP16AD and Gal4DBD lines. To exploit the superior specificity of split-Gal4 drivers at the ligand-expressing (sending) neurons, we replaced the LexA>LexAop system with the split-Gal4>UAS system, significantly enhancing the versatility of TRACT. Consequently, since split-Gal4>UAS is employed to express the ligand, we adopted the QF2>QUAS system for constructing the synthetic Notch proteolysis system. To specifically label astrocytes associated with targeted neuronal populations, we express the synthetic Notch proteolysis system selectively in astrocytes as the receiving cells and refer to this modified system as astro-TRACT (Figure 1A). Specific split-Gal4 drivers can be directly crossed to a single genotype harboring the remaining three genetic elements, providing a user-friendly and flexible approach to investigate astrocytes closely associated with defined neuronal circuits (Figure 1A).

Figure 1.

Adapted Notch-mediated genetic tool labels local astrocytes with high precision

(A) Schematic representation of sequence of events occurring with astro-TRACT tool. Left: extracellular CD19 ligand at the neuronal presynaptic site connects to extracellular anti-CD19 domain fused with a modified Notch receptor, which is located in the membrane of all astrocytes. Middle: the transcription factor QF2 is cleaved from the modified Notch receptor and translocates into the nucleus. Right: once in the nucleus, QF2 will bind to QUAS, thereby expressing green fluorescence in local astrocytes. Any effector in place of GFP can be used. The tool contains four transgenetic elements: (1) split Gal4 driver, (2) UAS-nSyb-CD19-Ollas, (3) R86E01-DSCP/alrm-nlg-synthetic_Notch-anti_CD19-QF2-V5, and (4) QUAS-effector.

(B–E) Immunostaining of astrocytes (green) and sending neuron (magenta) with astro-TRACT tool. NC82/brp staining (gray) for reference of whole brain structure. Sending neuronal circuit are KCs. Astrocyte labeling around MB lobes differs between (B) original version¹⁰ alrm-synthetic_Notch-anti_CD19-Gal4-V5 via P element (second chromosome) injection, and (C) modified alrm-nlg-synthetic_Notch-anti_CD19-Gal4-V5 injected into VK00027 docking site, (D and E) alrm-nlg-synthetic_Notch-anti_CD19-QF2-V5 injected into (D) VK00027 or (E) JK22C docking sites. Genotypes and docking sites are also summarized in Table S1. Scale bars, 50 μm.

Reporter expression in receiving astrocyte is affected by astrocyte-specific enhancer and insertion site

As we changed the transcriptional activator in the synthetic Notch proteolysis system from Gal4 in TRACT¹⁰ to QF2 in astro-TRACT, we compared reporter intensity and expression pattern resulting from using either system. While we found similar labeling results comparing Gal4 and QF2 to drive reporter expression (Figures 1B–1E and S1), both the choice of the enhancer driving transcriptional activator expression in astrocytes (alrm and R86E01) and the genomic insertion site of the associated transgene affected tool efficiency. Typically, we found that the R86E01 enhancer returns stronger GFP reporter expression compared to alrm. While in the original TRACT¹⁰ the astrocyte Gal4 transgenic element was integrated into the second chromosome via P element transposition, we here injected either into VK00027 (third chromosome) or JK22C (second chromosome) docking sites. The astrocyte signal around the MB became more specific with this adaptation, including the absence of GFP signal in the optic lobes (Figures 1B–1E andS1). We did observe unspecific neuronal labeling in the posterior lateral part of the central brain with the R86E01-Gal4 version injected into VK00027 (Figure S1A), but not with the alrm enhancer. Therefore, we injected the R86E01-QF2 version only into JK22C (Figure S1B), not VK00027. The combined changes to the original TRACT system have improved the specificity and usability of the tool in adult Drosophila brains.

Sensitivity of astro-TRACT is driven by the number of pre-synaptic sites of sending neuronal circuit

Beside testing the specificity of astro-TRACT, we asked how sensitive and reproducible the astrocyte labeling is. We tested astrocyte labeling with 18 split-Gal4 drivers that express neurons with varying numbers of pre-synaptic sites in different parts of the MB lobe compartments. Fourteen of the 18 drivers label at least one astrocyte in the MB region. The labeling in many of them is so sparse that exactly one astrocyte is labeled, either in both or even in one hemisphere only (Figures 2A–2L; Table S2). Also, we observed that processes of astrocytes typically stay on the same side of the brain as their soma, with the exception of two protocerebral anterior medial (PAM) dopamine neuron-specific drivers targeting γ3, 4, and 5 MB compartments (MB196B and MB042B). In those cases, labeled astrocytes project processes across the midline into the MB lobes of the other hemisphere. The chance of reproducing labeling across brains of the same genotype depends on the driver. While the reproducibility is 100% in 10 out of the 14 drivers, it is more variable for the remaining 4 drivers (Table S2).

Figure 2.

Number of pre-synapses in targeted neuronal cell type determines sensitivity of astro-TRACT tool

(A–L) Immunostaining of MB-local astrocytes (green) around different compartments of the γ (A–F), α’/β’ (G–I), and α/β (J–L) lobes (magenta). NC82/brp staining (gray) for reference of whole brain structure. Scale bars, 50 μm. Labeling results for all tested MB driver lines are summarized in Table S2.

(M) Comparison of pre-synapse number across MB driver lines targeting KC (red shades), MBON (orange shades), PPL1 (green) and PAM (blue shades) neurons, used in this study. Size of shapes indicates number of cells matched in connectome. Underlying table to plot can be found in Table S3.

While all 18 drivers project axons into MB lobe compartments, they originate from different cell types: PAM and PPL1 dopaminergic neurons, Kenyon cells (KCs), and mushroom body output neurons(MBONs) and drive reporter expression in a varying number of neurons. For example, the two MBON drivers used in this study target just one neuron which project mostly post-synaptic but also few pre-synaptic sites into MB compartments.¹² We found that successful astrocyte labeling depends on the chosen sending neuronal cell type. While all drivers targeting PAM dopaminergic neurons and KCs were successful, all drivers targeting PPL1 dopaminergic neurons and MBONs were not (Figure 2M; Table S2). In the latter cases, instead of astrocyte labeling, we frequently noticed unspecific neuronal, but never unspecific glial signal. Generally, chances of unspecific neuronal labeling increase with higher sparsity of the sending neuronal circuit and heavily depend on the driver used.

We hypothesized that the reason for this discrepancy may be that the number of pre-synapses in a given MB compartment to catalyze the Notch mechanism of astro-TRACT differs between cell types. To test this quantitatively, we estimated the number of pre-synapses per MB compartment based on public hemibrain connectome data^13,14 detailing synaptic connectivity across the fly brain. We found that those neuronal driver lines that result in labeled astrocytes have a higher estimated number of pre-synaptic sites in MB lobe compartments (Figure 2M; Table S3). Therefore, a certain level of pre-synapses appears to be required in an MB compartment to catalyze the TRACT mechanism and label local astrocytes (Figure 2M; Table S3). In the future, the accuracy of this analysis can be improved by increasing the number of cells matched to drivers as more traced neurons become available in the connectome. Still, estimating synaptic density may be a useful initial sensitivity predictor for different neuronal circuits.

To probe the sensitivity of astro-TRACT in other circuits, we tested two different drivers targeting the R5 neuron subtype of the ellipsoid body (EB). Similar to other sparse cell populations (MBON and PPL1 in Figure 2M; Tables S2 and S3), we found that the labeling was unsuccessful for one driver of R5 cells (Figure S2B). Interestingly, we did observe consistent labeling when using the other R5 driver, albeit with unspecific neuronal labeling (Figure S2A). This discrepancy may arise because these drivers differ in either expression strength or the subpopulation of R5 cells that they target or both.

To test our hypothesis that astrocyte labeling with astro-TRACT depends on synaptic density of the sending circuit, we compared astrocyte labeling with astro-TRACT across developmental stages, specifically late larval, late pupal, and adult stages (Figures S3A–S3C). We observed GFP+ astrocytes only in adults, with earlier stages showing only nonspecific signal. This indicates that either astrocyte labeling cannot occur before sufficient synapse maturation, that CD19-synthetic Notch binding and subsequent QF2 translocation require time into adulthood, or that both factors contribute.

Local astrocytes around MB connect to EB

Next, we were interested in applying astro-TRACT to examine individual astrocyte morphology around the MB and MB-adjacent regions like the EB. In addition to their close proximity, the neuropil share similar functions, such as sleep behavior regulation. Yet according to the Drosophila connectome, there are no neuronal connections between these structures.^15,16,17

We asked whether instead of neuronal connections, astrocytes may play a role in bridging the two regions. To address this question, we combined astro-TRACT with a pan-MB driver (MB247-LexA) and a multi-color mosaic system. MCFO (MultiColor FlpOut) is a technique to stochastically label individual cells with three tags.¹⁸ The integration of astro-TRACT and MCFO allowed us to visualize a sparse number of astrocytes that are located specifically around the MB with different colors (Figures 3C–3E). When combining the MB-local astrocyte signal of the three MCFO tags, almost the entire circular structure of the EB forms (Figures 3A and 3B), suggesting that MB-local astrocytes do connect to the EB ring. Using astro-TRACT combined with MCFO and an MB-specific driver, we found that sparsely labeled MB-local astrocytes collectively trace out the EB ring, suggesting that astrocytes may provide a morphological bridge between these neighboring brain regions with shared functions.

Figure 3.

Local astrocytes of MB connect to EB

(A and B) Immunostaining of MCFO tags HA (cyan), FLAG (yellow), and V5 (magenta) in brains expressing both astro-TRACT and MCFO tools under pan-MB promoter. NC82/brp staining (gray) for reference of whole brain structure. Composite images of all four channels combined. The maximum intensity projection includes all z planes with the EB structure present. Scale bars, 50 μm. (B) Arrow indicates astrocytes that connect MB and EB.

(C–E) Maximum intensity projections of separate tags showing labeling of individual astrocytes. (C) MB and EB structure are roughly indicated by dotted lines.

Astrocytes specifically bridge medial MB compartments of the horizontal lobes with the EB

Next, we asked whether the MB-astrocyte-EB connection is specific to certain MB compartments. Probing this connection required an increased resolution to visualize individual astrocytes. For this purpose, we analyzed single astrocyte labeling with astro-TRACT across 14 split-Gal4 drivers targeting the pre-synapses in different MB compartments (Table S2).

We found that astrocytes connect a respective MB compartment to the EB for 6 of the 14 driver lines tested (Table S2; Figures 2B, 2C, 2H–2K, and 4B–4D). Typically, astrocytes connecting MB and EB are associated with β′2, β2, γ4, and γ5 and notably often cover more than one MB compartment at the same time. Interestingly, these astrocytes are not limited to the α/β lobe located closest to the EB, but they extend to the anterior γ lobe (Table S2). The labeled astrocytes that do not connect to the EB are on the lateral side of the MB lobes. We also observed that astrocytes typically project processes within the same hemisphere of their soma’s location, except for two PAM neuron-specific drivers labeling the γ3, γ4, and γ5 MB compartments (MB196B and MB042B), in which cases the astrocytes span across the midline.

Figure 4.

Local astrocytes of MB connect to EB ring neurons

(A) Illustration of MB (magenta) and EB (cyan) positions: MB is located anterior to EB and its medial compartments are closest to the EB. Zoomed in (left) and in relation to whole fly brain (right). Right image was extracted from flywire.ai.

(B) Maximum intensity projection of immunostained MB β′2 compartment (magenta) and EB ring structures of brains expressing astro-TRACT system under MB418B-Gal4 promoter. Astrocyte (green) innervates the EB ring neuropil. Dashed line outlines EB ring neuron structure. Scale bars, 20 μm.

(C and D) Two angles of 3D reconstruction of z stack from (B) showing how local astrocyte connects both structures.

(E) 4 cell IDs annotated as astrocytes from the flywire connectome connect MB and EB structures.

(F and G) Two example cell IDs showing astrocyte morphology and connection between MB and EB neuropils.

Next, we asked whether there are any cells in the recently published connectome dataset^14,15 that are classified as non-neuronal, exhibit an astrocyte-typical shape and connect the MB and EB. Currently, we identified four astrocytes that indeed connect these two regions. This number will likely increase as cell reconstructions are ongoing. In accordance with our immunostaining results, astrocyte cell bodies are located medially and attach to MB compartments with laterally projecting processes (Figures 4E–4G). Thus, single-astrocyte analyses reveal that specifically medial MB compartments (β′2, β2, γ4, and γ5) are bridged to the EB by astrocytes, a pattern supported by connectome data.

Discussion

Understanding the roles that astrocytes play in signal processing and animal behavior requires targeting specific astrocyte subpopulations. In this study, we examined and optimized the genetic tool TRACT to label local astrocytes associated with user-defined Drosophila neuronal circuits in adults. The success of astro-TRACT, i.e., the specific and exclusive labeling of the desired astrocyte subpopulation relies on multiple factors, including the enhancer and injection site used for the receiving astrocyte transgene and sufficient synaptic density in the sending neuronal circuit. Our comparison between KC, PAM, MBON, and PPL1 sending circuits has shown that higher synaptic density driven by the neuronal driver maximized the success rate of labeling local astrocytes. To enhance labeling specificity, we switched the binary systems of TRACT, allowing its combination with highly specific neuronal split-Gal4 drivers. Furthermore, the extensive Gal4 driver collection creates high flexibility for interested users. In addition, in the future astro-TRACT can be adapted to other glial cell classes like ensheathing and perineurial glia by replacing the astrocyte enhancer, providing versatility in labeling experiments.

The tool’s greatest flexibility is provided by its plug-in design of astrocyte effectors. The available library of effectors under UAS control is vast, including calcium activity indicators, gene knockdown with RNAi and CRISPR-Cas9 for functional studies, and GFP or other tag reporters for morphology characterization. The combination of these available tools with astro-TRACT allows users to adapt the tool for their specific research purposes. Furthermore, the labeling strategy is based on physical proximity of astrocyte to synapse rather than astrocyte-specific secretion molecules, making it a valuable tool for studying unknown astrocyte-neuron signaling mechanisms. By using functional indicators under UAS control in local astrocytes, astro-TRACT provides opportunities to unravel unknown astrocyte-neuron signaling pathways and to explore how these interactions may vary across different brain regions.

There are two alternatives to label individual fly astrocytes: single clone labeling with MARCM (mosaic analysis with a repressible cell marker)¹⁹ and MCFO (multi-color Flp out).¹⁸ In contrast to astro-TRACT, MARCM labels cells randomly and inconsistently between experiments. The multi-color mosaic technique MCFO can only label a cell type as specific as the driver used with it. In case of astrocytes, this means it will be expressed in astrocytes across the brain, which complicates analyses of individual astrocytes. Astro-TRACT provides higher reproducibility and specificity compared to these two existing tools, because local astrocytes of a chosen specific neuronal circuit are labeled.

After optimizing astro-TRACT, we used the tool to examine astrocyte projections from MB compartments to the close-by EB region. We found that astrocytes that surround compartments γ4, γ5, β′2, and β2 connect to the EB. We also identified similarly positioned cells in the FlyWire connectome. We propose that this MB-EB connection by astrocytes may facilitate efficient communication between these regions. As the computational center for learning and memory in the fly brain, the MB processes sensory and cognitive information.¹² Its activity is accompanied by an accumulation of sleep need, indicated by increases of pre-synaptic proteins in response to sleep deprivation.²⁰ Therefore, astrocytes connecting MB to EB may transfer sleep pressure that accumulates in an experience and activity-dependent manner in the MB²¹ to the sleep drive regulating R5 neurons of the EB.²² Astro-TRACT will be a valuable tool to test this hypothesis in the future.

In summary, we adapted the genetic tool TRACT to be used with specific split-Gal4 drivers targeting local astrocytes and characterized their expression specificity, sensitivity, and reproducibility in the fly MB. In the future, this tool can provide the resolution necessary to identify novel signal transmission mechanisms between astrocytes and neurons.

Limitations of the study

The labeling sensitivity of astro-TRACT may vary across neuronal circuits. We have observed that astrocytes are labeled consistently for certain cell types (KC and PAM) regardless of the driver used, but not for others (EB-R5). Unsuccessful astrocyte labeling is typically also associated with nonspecific neuronal labeling in other brain regions.

Furthermore, astro-TRACT combines two binary expression systems. Using Gal4/UAS to define the neuronal circuit provides access to the extensive library of highly specific split-Gal4 driver lines. Reserving the Gal4/UAS system for neurons demands the use of an alternative system, QF2/QUAS, to drive reporter expression in local astrocytes. Although, the UAS system offers the most extensive collection of reporter lines, UAS-based constructs can be adapted to QUAS relatively easily.

Resource availability

Lead contact

Requests for further information and resources should be directed to and will be fulfilled by the lead contact, Dr. Sha Liu (sha.liu@kuleuven.be).

Materials availability

The transgenic fly lines generated for this study will be deposited at the Bloomington Drosophila Stock Center. Plasmids constructed in this study will be deposited with Addgene. These reagents are also available upon request from the lead contact.

Data and code availability

• •All data reported in this paper will be shared by the lead contact upon request.
• •The code generated for analysis and visualization of synapse number per ROI across driver lines is available on GitHub (https://github.com/joanadopp/astroTRACT_paper). An archival version is available: https://doi.org/10.5281/zenodo.18338021.
• •Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

Acknowledgments

Imaging was supported by the light microscopy expertise unit of VIB-KU Leuven Center for Neuroscience. We thank Carlos Lois (Caltech) for sharing plasmids and fly stocks. We also thank the Bloomington Drosophila Stock Center for providing fly stocks. We thank the Princeton FlyWire team and members of the Murthy and Seung labs, as well as members of the Allen Institute for Brain Science, for the development and maintenance of FlyWire (supported by BRAIN Initiative grants MH117815 and NS126935 to Murthy and Seung). We also acknowledge members of the Princeton FlyWire team and the FlyWire consortium for neuron proofreading and annotation. Special thanks to Marissa Sorek for support and Nikitas Serafetinidis, Joseph Hsu, Arti Yadav, and Ryan Willie from the FlyWire Consortium for contributing >10% of the edits of cell reconstructions with cell_IDs 720575940617224498, 720575940607160753, 720575940627370277, and 720575940648849540. We thank Dr. Xiaojun Xie (Zhejiang University) for helpful comments on this manuscript. This work was funded by a starting grant of the European Research Council (#758580) and project grant of the Reseaerch Foundation – Flanders (G074923N) to S.L. J.D. held a PhD Fellowship of the Research Foundation - Flanders (#11D8820N).

Author contributions

Conceptualization, J.D. and S.L.; methodology, J.D., L.N., and J.Y.; investigation, J.D., S. Martens, F.H., S. Mastroianni, and A.M.; writing – original draft, J.D.; writing – review & editing, J.D., F.H., S. Mastroianni, and S.L.; funding acquisition, J.D. and S.L.; resources, S.L.; supervision, J.D. and S.L.

Declaration of interests

The authors declare no competing interests.

STAR★Methods

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Anti-GFP (rabbit)	Invitrogen	Cat # A11122; RRID: AB_221569
Anti-Ollas (rat)	Novus Biologicals	Cat # NBP1-06713; RRID:AB_1625979
Anti-NC82 (mouse)	Developmental Studies Hybridoma Bank	Cat # AB_2314866; RRID:AB_2314866
Anti-Flag (rat)	Novus Biologicals	Cat # NBP1-06712; RRID:AB_1625981
Anti-HA (rabbit)	Roche	Cat # 11867423001; RRID:AB_390918
Secondary anti-rabbit 488	Invitrogen	Cat # A11008; RRID:AB_143165
Secondary anti-rat 568	Invitrogen	Cat # A11077; RRID:AB_141874
Secondary anti-mouse 647	Invitrogen	Cat # A21235; RRID:AB_2535804
Secondary anti-mouse 488	Invitrogen	Cat # A11029; RRID:AB_2534088
Secondary anti-rabbit 594	Jackson Immuno Research	Cat # 711-585-152; RRID:AB_2340621
Secondary anti-rat 647	Invitrogen	Cat # A21247; RRID:AB_141778
DL550 anti-V5 (mouse)	BioRad	Cat # MCA1360D550GA; RRID:AB_2687576
Critical commercial assays
Gibson Assembly Master Mix	NEB	#E2611
Deposited data
HemiBrain connectome	Scheffer et al.13	https://neuprint.janelia.org/?dataset=hemibrain:v1.2.1
FAFB-FlyWire connectome	Schlegel et al.14; Dorkenwald et al.15	https://ngl.flywire.ai/; https://codex.flywire.ai/
Experimental models: Organisms/strains
D. melanogaster: w[1118]	Bloomington Drosophila Stock Center	BDSC: 5905
D. melanogaster: MB247-Gal4: w[∗]; P{w[+m∗] = Mef2-GAL4.247}3	Bloomington Drosophila Stock Center	BDSC: 50742
D. melanogaster: MB247-LexA: w[∗]; l(2)∗[∗]/CyO; P{w[+mC] = Mef2-LexA-VP16.247}3/TM6B, Tb[1]	Bloomington Drosophila Stock Center	BDSC: 606105
D. melanogaster: R58H05-GAL4: w[1118]; P{y[+t7.7] w[+mC] = GMR58H05-GAL4}attP2	Bloomington Drosophila Stock Center	BDSC: 39198
D. melanogaster: UAS-mCD8:GFP: P{w[+mC] = UAS-salm.K}X, w[∗]; P{w[+mC] = UAS-mCD8:GFP.L}LL5	Bloomington Drosophila Stock Center	BDSC: 29715
D. melanogaster: QUAS-mCD8-GFP: y[1] w[1118]; P{w[+mC] = QUAS-mCD8-GFP.P}5B/TM6B, Tb[1]	Bloomington Drosophila Stock Center	BDSC: 30003
D. melanogaster: MCFO-1: w[1118] P{y[+t7.7] w[+mC] = hs-FLPG5.PEST}attP3; PBac{y[+mDint2] w[+mC] = 10xUAS(FRT.stop)myr:smGdP-HA}VK00005 P{y[+t7.7] w[+mC] = 10xUAS(FRT.stop)myr:smGdP-V5-THS-10xUAS(FRT.stop)myr:smGdP-FLAG}su(Hw)attP1	Bloomington Drosophila Stock Center	BDSC: 64085
D. melanogaster: MB010B: w[1118]; P{y[+t7.7] w[+mC] = R13F02-p65.AD}attP40; P{y[+t7.7] w[+mC] = R52H09-GAL4.DBD}attP2	Bloomington Drosophila Stock Center	BDSC: 68293
D. melanogaster: MB025B: w[1118]; P{y[+t7.7] w[+mC] = R24E12-p65.AD}attP40; P{y[+t7.7] w[+mC] = R52H01-GAL4.DBD}attP2	Bloomington Drosophila Stock Center	BDSC: 68299
D. melanogaster: MB112C: w[1118]; P{y[+t7.7] w[+mC] = R13F04-GAL4.DBD}attP2 PBac{y[+mDint2] w[+mC] = R93D10-p65.AD}VK00027	Bloomington Drosophila Stock Center	BDSC: 68263
D. melanogaster: MB188B: w[1118]; P{y[+t7.7] w[+mC] = R58E02-p65.AD}attP40/CyO; P{y[+t7.7] w[+mC] = R11A03-GAL4.DBD}attP2/TM6B, Tb[1]	Bloomington Drosophila Stock Center	BDSC: 68268
D. melanogaster: MB196B: w[1118]; P{y[+t7.7] w[+mC] = R58E02-p65.AD}attP40/CyO; P{y[+t7.7] w[+mC] = R36B06-GAL4.DBD}attP2	Bloomington Drosophila Stock Center	BDSC: 68271
D. melanogaster: MB298B: w[1118]; P{y[+t7.7] w[+mC] = R53C03-p65.AD}attP40; P{y[+t7.7] w[+mC] = R24E12-GAL4.DBD}attP2/TM6B, Tb[1]	Bloomington Drosophila Stock Center	BDSC: 68309
D. melanogaster: MB312B: w[1118]; P{y[+t7.7] w[+mC] = R58E02-p65.AD}attP40/CyO; P{y[+t7.7] w[+mC] = R10G03-GAL4.DBD}attP2	Bloomington Drosophila Stock Center	BDSC: 68314
D. melanogaster: MB315C: w[1118]; P{y[+t7.7] w[+mC] = R48H11-GAL4.DBD}attP2 PBac{y[+mDint2] w[+mC] = R58E02-p65.AD}VK00027	Bloomington Drosophila Stock Center	BDSC: 68316
D. melanogaster: MB418B: w[1118]; P{y[+t7.7] w[+mC] = R26E07-p65.AD}attP40/CyO; P{y[+t7.7] w[+mC] = R30F02-GAL4.DBD}attP2	Bloomington Drosophila Stock Center	BDSC: 68322
D. melanogaster: MB441B: w[1118]; P{y[+t7.7] w[+mC] = R30G08-p65.AD}attP40; P{y[+t7.7] w[+mC] = R48B03-GAL4.DBD}attP2	Bloomington Drosophila Stock Center	BDSC: 68251
D. melanogaster: MB504B: w[1118]; P{y[+t7.7] w[+mC] = R52H03-p65.AD}attP40; P{y[+t7.7] w[+mC] = ple-GAL4.DBD}attP2	Bloomington Drosophila Stock Center	BDSC: 68329
D. melanogaster: MB607B: w[1118]; P{y[+t7.7] w[+mC] = R19B03-p65.AD}attP40; P{y[+t7.7] w[+mC] = R39A11-GAL4.DBD}attP2	Bloomington Drosophila Stock Center	BDSC: 68256
D. melanogaster: MB042B: w[1118]; P{y[+t7.7] w[+mC] = R58E02-p65.AD}attP40/CyO; P{y[+t7.7] w[+mC] = R22E04-GAL4.DBD}attP2	Bloomington Drosophila Stock Center	BDSC: 68303
D. melanogaster: MB009B: w[1118]; P{y[+t7.7] w[+mC] = R13F02-p65.AD}attP40/CyO; P{y[+t7.7] w[+mC] = R45H04-GAL4.DBD}attP2	Bloomington Drosophila Stock Center	BDSC: 68292
D. melanogaster: MB320C: w[1118]; P{y[+t7.7] w[+mC] = R22B12-GAL4.DBD}attP2 PBac{y[+mDint2] w[+mC] = ple-p65.AD}VK00027	Bloomington Drosophila Stock Center	BDSC: 68253
D. melanogaster: MB463B: w[1118]; P{y[+t7.7] w[+mC] = R35B12-p65.AD}attP40; P{y[+t7.7] w[+mC] = R34A03-GAL4.DBD}attP2/TM6B, Tb[1]	Bloomington Drosophila Stock Center	BDSC: 68370
D. melanogaster: MB185B: w[1118]; P{y[+t7.7] w[+mC] = R52H09-p65.AD}attP40; P{y[+t7.7] w[+mC] = R18F09-GAL4.DBD}attP2	Bloomington Drosophila Stock Center	BDSC: 68267
D. melanogaster: MB008B: w[1118]; P{y[+t7.7] w[+mC] = R13F02-p65.AD}attP40; P{y[+t7.7] w[+mC] = R44E04-GAL4.DBD}attP2	Bloomington Drosophila Stock Center	BDSC: 68291
D. melanogaster: MB371B: w[1118]; P{y[+t7.7] w[+mC] = R13F02-p65.AD}attP40; P{y[+t7.7] w[+mC] = R85D07-GAL4.DBD}attP2	Bloomington Drosophila Stock Center	BDSC: 68383
D. melanogaster: R46C03-AD (VK00033)	Liu lab	N/A
D. melanogaster: VT45101-DBD: w[1118]; P{y[+t7.7] w[+mC] = VT045101-GAL4.DBD}attP2	Bloomington Drosophila Stock Center	BDSC: 73345
D. melanogaster: LexAop-nSyb-CD19-Ollas (attP2)	Gift of C. Lois	Huang et al.11
D. melanogaster: alrm-synthetic_Notch-anti-CD19-Gal4-V5	Gift of C. Lois	Huang et al.10
D. melanogaster: alrm-nlg-synthetic_Notch-anti_CD19-Gal4-V5 (VK00027)	This paper	N/A
D. melanogaster: alrm-nlg-synthetic_Notch-anti_CD19-QF2-V5 (VK00027)	This paper	N/A
D. melanogaster: alrm-nlg-synthetic_Notch-anti_CD19- QF2-V5 (JK22C)	This paper	N/A
D. melanogaster: R86E01-nlg-synthetic_Notch-anti_CD19-Gal4-V5 (VK00027)	This paper	N/A
D. melanogaster: R86E01-nlg-synthetic_Notch-anti_CD19-QF2-V5 (JK22C)	This paper	N/A
D. melanogaster: 10xUAS-nSyb-CD19-Ollas (JK22C)	This paper	N/A
D. melanogaster: 10xUAS-nSyb-CD19-Ollas (JK73A)	This paper	N/A
Oligonucleotides
Primer: alrm-F: 5′ GGA ACT AGG CTA GCA CTA CGC ACA GAT GTG GTC ATC TGA ATA GG 3′	This paper	N/A
Primer: alrm-R: 5′ CAC CCA TGG TGG AAT TCT AGT GGC GAT CCT TTC GCT CGG GAG C 3′	This paper	N/A
Primer: R86E01-F: 5′ GAA AAT GCT TGG ATT TCA CTG GAA CTA GGC TAG CCA CAT AAT ACT CTA CAG GGC ATC CAC 3′	This paper	N/A
Primer: R86E01-R: 5′ GAT CCC CGG GCG AGC TCG GAT TGT GAA GCC CCA AGA GTA C 3′	This paper	N/A
Primer: DSCP-F: 5′ GTA CTC TTG GGG CTT CAC AAT CCG AGC TCG CCC GGG GAT CG 3′	This paper	N/A
Primer: DSCP-R: 5′ CCC AGG AGC TGG GTA GGG ACA CCC ATG GTG GAA TTC GTT TGG TAT GCG TCT TGT GAT TC 3′	This paper	N/A
Primer: QF2-F: 5′ GTA CAT ATT CAG GAA ATC TCA GTC GGC AGC TCT AGA CCA CCC AAG CGC AAA ACG CTT AAC 3′	This paper	N/A
Primer: QF2-R: 5′ CTT TAG TCG ACG GTA TCG ATA GAC GGG CGC GCC TCA TCA CTG TTC GTA TGT ATT AAT GTC 3′	This paper	N/A
Primer: nlg-synthetic_Notch-anti_CD19-F: 5′ GGT GGA ACT GCC ACC ATC ATC GTC GAC GGG CCA GAC GTC 3′	This paper	N/A
Primer: nlg-synthetic_Notch-anti_CD19-R: 5′ GCT TGG GTG GTC TAG AGC TGC CGA CTG AGA TTT CCT GAA TAT GTA CAC GTT TCT TGC CGC 3′	This paper	N/A
Recombinant DNA
nSyb-nlg-SNTG4-V5	Gift of C. Lois	Huang et al.11
13xLexAop2-IVS-nsyb:CD19	Gift of C. Lois	Huang et al.11
pCasper4-QF#7-hsp70	Addgene	# 46135
pJFRC81-10xUAS-IVS-Syn21-GFP-P10	Addgene	# 36432
pALRM-nlg-synthetic_Notch-anti_CD19-Gal4-V5	This paper	N/A
pR86E01-DSCP-nlg-synthetic_Notch-anti_CD19-Gal4-V5	This paper	N/A
pALRM-nlg-synthetic_Notch-anti_CD19-QF2-V5	This paper	N/A
pR86E01-DSCP-nlg-synthetic_Notch-anti_CD19-QF2-V5	This paper	N/A
p10xUAS-IVS-nsyb-CD19-Ollas	This paper	N/A
Software and algorithms
Fiji	Schindelin et al.23	https://fiji.sc/; RRID: SCR_002285
Imaris version 10.0.1	Oxford Instruments, https://imaris.oxinst.com/	RRID: SCR_007370
R version 4.3.3	R Core Team24	RRID:SCR_001905
natverse (R) version 1.10.4	Bates et al.25	https://natverse.org/natverse/
neuprintr (R)	Bates et al.25	https://natverse.org/neuprintr/
neuronbridger (R)	Clements et al.26; Otsuna et al.27	https://github.com/natverse/neuronbridger
ggplot2 version 3.5.1	Wickham28	RRID: SCR 014601
Code for synapse number per ROI calculation and plotting	This paper	https://github.com/joanadopp/astroTRACT_paper

Experimental model and study participant details

Fly husbandry

Flies were reared at 25°C on a standard cornmeal-agar diet under a 12:12 h light-dark cycle, except for MCFO experiments. Flies were aged 4–7 days post-eclosion before dissections. Both female and male flies were used.

Method details

Generation of Drosophila lines

alrm (or R86E01-DSCP)-nlg-synthetic_Notch-anti_CD19-Gal4-V5

The enhancer nSyb of the nSyb-nlg-SNTG4-V5 construct (gift from Lois lab, Caltech) was cut out with Nhe1 and EcoR1. Two lines, one with the 5069 bp long alrm enhancer²⁹ and another one with the 3140 + 155 bp R86E01-DSCP enhancer (Flylight) were constructed by cloning the respective enhancer fragment with overlapping homology arms into the nlg-SNTG4-V5 backbone by Gibson assembly (NEB).

alrm (or R86E01-DSCP)-nlg-synthetic_Notch-anti_CD19-QF2

Another two lines were generated to contain the QF2 transcription factor instead of Gal4. The transcription factor QF2 was isolated from pCasper4-QF#7-hsp70 (Addgene #46135). The backbone alrm (or R86E01-DSCP)-nlg-synthetic_Notch-anti_CD19 was cut with Aat2 and Asc1. The 530 bp fragment of the part of the nlg-synthetic_Notch-anti_CD19 that was cut away from the backbone was amplified from nSyb-nlg-SNTG4-V5 and assembled with the amplified QF2 transcription factor with overlapping homology between inserts and backbone by Gibson assembly (NEB).

10xUAS-IVS-nSyb-CD19

The 13xLexAop2 fragment of the construct 13xLexAop2-IVS-nsybCD19 (gift from Carlos Lois, Caltech) was removed and replaced by the 10xUAS fragment isolated from pJFRC81-10xUAS-IVS-Syn21-GFP-P10 (Addgene #36432) with Bgl2 and Hind3 by conventional T4 DNA ligase-mediated cloning (NEB).

The transgenes were injected into fly embryos in-house in the iso31 background by site-directed PhiC31-mediated insertion into VK00027 (alrm-nlg-synthetic_Notch-anti_CD19-Gal4-V5, alrm-nlg-synthetic_Notch-anti_CD19-QF2-V5, R86E01-DSCP-nlg-synthetic_Notch-anti_CD19-Gal4-V5) and JK73A (10xUAS-IVS-nSyb-CD19-Ollas) on the third chromosome and JK22C (alrm-nlg-synthetic_Notch-anti_CD19-QF2-V5, R86E01-DSCP-nlg-synthetic_Notch-anti_CD19-QF2-V5, 10xUAS-IVS-nSyb-CD19-Ollas) on the second chromosome.

After injection, one of multiple transformants was selected for the highest labeling specificity and efficiency. The chosen transformant was outcrossed 5 times to iso31. Subsequently, multiple genetic recombinations were performed. Various combinations of transgenes and injection sites were recombined to optimize the labeling sensitivity and specificity and tool versatility.

Multi-color Flp-Out (MCFO) analyses

MCFO is a technique in which three different tags under UAS control are by default silenced by FRT-flanked transcriptional terminators.¹⁸ Heat-shock induced FLPase expression randomly removes the terminators in individual cells that also express a GAL4 driver. This results in a mosaic of differently colored cells within the user-defined cell type. Flies with genotype HsFlpG5.Pest; alrm-synthetic_Notch-anti_CD19-Gal4; UAS-McFlip (UAS-STOP-smGFP-HA, UAS-STOP-smGFP-V5, UAS-STOP-smGFP- FLAG), MB247-LexA, LexAop-nSyb-CD19-Ollas were raised at 18°C. Adult 2-4 days-old flies were heat-shocked by being placed into a 37°C water bath for 5 min. Subsequently, flies recovered for 24 h at 25°C before dissection.

Immunohistochemistry on whole-mount adult Drosophila brain

GFP, Ollas tag and NC82

After dissection in Schneider’s medium, fly brains were fixed in 2% paraformaldehyde for 55 min on a nutator. Next, brains were washed nutating at least four times for 15 min each in 0.5% PBST. Following that, the brains were blocked for 1.5 h at room temperature in 5% Normal Goat Serum [Jackson Immuno Research 005-000-121] in 0.5% PBST. Then they were incubated in primary antibodies overnight at 4°C (Rabbit anti GFP 1:500 [Invitrogen A11122], Rat anti Ollas 1:200 [Novus Biologicals NBP1-06713], Mouse anti NC82 1:25 [DSHB]). The next day, brains were washed at least four times for 15 min each in 0.5% PBST. The secondary antibody (anti-Rabbit 488 1:1000 [Invitrogen A11008], anti-Rat 568 1:500 [Invitrogen A11077], anti-Mouse 647 1:300 [Invitrogen A21235]) incubation followed overnight at 4°C. Finally, the brains were washed at least four times for 15 min each in 0.5% PBST and left to incubate overnight at 4°C in Vectashield before mounting.

MCFO

The protocol for mosaic labeling was based on the ‘FlyLight Protocol – MCFO IHC for Adult Drosophila CNS’ and followed the same protocol as above with the following exceptions: The primary (Rat anti Flag 1:200 [Novus Biologicals NBP1-06712], Rabbit anti HA 1:300 [Roche 11867423001], Mouse anti NC82 1:25) and secondary antibodies (anti-Mouse 488 1:400 [Invitrogen A11029], anti-Rabbit 594 1:500 [Jackson Immuno Research 711-585-152], anti-Rat 647 1:300 [Invitrogen A21247]) were incubated for four hours at room temperature before 2 or 3–4 overnights at 4°C for primary and secondary, respectively. After the post-secondary antibody washes, an incubation in 5% Normal Mouse Serum (NMS) [Jackson Immuno Research 015-000-120] in 0.5% PBST followed for 1.5 h on a nutator. Subsequently, brains were incubated in direct label DL550 Mouse anti-V5 antibody (1:500) [BioRad MCA1360D550GA] in 5% NMS in 0.5% PBST for 4 h at room temperature and another one overnight at 4°C. Finally, brains were washed at least 4 times for 15 min each in 0.5% PBST at room temperature and left to incubate overnight at 4°C in Vectashield.

Image acquisition and analysis

Mounted brains were imaged using either a Zeiss Airyscan 880 or a Nikon NiE A1R confocal microscope. Images were processed in Fiji ImageJ (NIH). 3D reconstruction, segmentation and visualizations were generated with Imaris (Bitplane, version 10.0.1).

Quantification and statistical analysis

Connectome data analyses

Estimates of synaptic density by driver line were analyzed using the natverse libraries (v 1.10.4)²⁵ in R (v 4.3.3). Cell IDs in the hemibrain:v1.2.1 dataset¹³ were matched by driver with the neuronbridge_line_contents function of the neuronbridger package (v 2.2.0). For 11 of the 15 MB-projecting drivers, we were able to map connectome cell IDs. We then extracted the number of pre-synapses per hemisphere, MB compartment and cell ID and averaged this number across all cell IDs by driver. Then we multiplied this average to the number of cells of the cell type targeted by the driver, estimated based on light microscopy images,^12,30,31 arriving at an estimated number of pre-synapses per compartment and driver. Plotting the estimated synapse number by MB compartment was performed with ggplot2 (v 3.5.1).

To visualize putative astrocytes connecting MB and EB in the connectome, we scanned 141 cells flagged as ‘not_a_neuron’ and centrally located in the brain that are listed in the Supplemental Information of¹⁴. We identified 4 such cell_IDs (720575940617224498, 720575940607160753, 720575940627370277, 720575940648849540). 3D cell visualizations were directly extracted from codex.flywire.ai.

Published: March 13, 2026