TRACR: an anterograde transneuronal tracing system for genetic access across synapses and longitudinal circuit analysis

Following neural signals as they converge onto and diverge from individual neurons is central to understanding circuit function and disease-related dysfunction. However, existing approaches to map connected structural reconstruction or functional recordings, or rely on cytotoxic viral tracers. To address these limitations, we adapted synthetic Notch designs to create TRanssynaptic Anterograde Circuit Readout (TRACR). Binding of the engineered ligand- receptor across synapses induces reporter transcription, enabling characterization and manipulation of postsynaptic neurons. By applying TRACR at multiple synapses in the mouse visual system, we show that TRACR labels postsynaptic partners of sensory neurons, long- range projections and local inhibitory interneurons. TRACR provides segregated genetic access to pre- and postsynaptic populations for expression of markers, sensors, or effectors. Finally, TRACR signaling is reversible and fails to activate when synapses are absent or disrupted. TRACR is an accessible, AAV-deliverable transneuronal reporting tool for longitudinal analysis of circuit assembly, organization, and degeneration. .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 8, 2026. ; https://doi.org/10.64898/2026.02.08.704659doi: bioRxiv preprint 3

A central goal of neuroscience is to explain how neural circuits encode, and how they change in disease states. Yet determining which neurons are synaptically-connected within a circuit remains technically challenging. Functional approaches, such as electrophysiology or optogenetic recordings, can demonstrate connectivity between individual neurons or populations, but are difficult to integrate with circuit-wide anatomical analyses. Structural approaches, including immunohistochemistry and electron microscopy, can identify synapse location and synaptic abnormalities, but not which neurons are connected within an intact circuit. An ideal circuit mapping tool would identify the connected cell types, work across diverse types of synapses in local and long-range circuits, and allow genetic access to both partners for molecular profiling and functional manipulations. As circuits remodel during development and disease, a tracing strategy designed to capture connectivity changes over time would also open the door to studying circuit assembly, degeneration, and repair. Current viral and intercellular strategies can link neural identity to connectivity, but their performance varies across synapse types, pathways, and experimental timelines. In mammals, the synapse crossing property of neurotropic viruses has been leveraged for transsynaptic tracing. For example, the glycoprotein-deleted rabies virus (RVdG) can be used to deliver tools to mark, monitor, or manipulate a starter neuron’s monosynaptic inputs, but are highly cytotoxic (Wickersham, Finke, et al. 2007; Wickersham, Lyon, et al. 2007; Osakada and Callaway 2013). Engineered HSV, VSV, yellow fever and AAV1 can anterogradely trace postsynaptic partners from a given neural type but are often limited by undesired retrograde spread, incomplete genetic access to starter and/or target populations, cytotoxicity, and biosafety or regulatory constraints (Lo and Anderson 2011; Beier et al. 2013; Zingg et al. 2017; Zingg et al. 2020; Li et al. 2021; Xiong et al. 2022; Bouin et al. 2024; Jin et al. 2024). Tools that exploit neurotransmission are promising (e.g., WGA, ATLAS, BAcTrace) but can be limited to certain synapse types and variable across circuits (Cachero et al. 2020; Tsai et al. 2022; Rivera et al. 2025). Importantly, most existing approaches—viral and non-viral—still yield largely static snapshots via material transfer or permanent reporter induction, limiting the ability to track connectivity changes when circuits remodel. To address these gaps, we developed a transneuronal tracing system based on synthetic Notch signaling (synNotch) that converts synaptic contact into a customizable and reversible genetic readout. SynNotch receptors retain the core Notch regulatory and transmembrane domains but .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 8, 2026. ; https://doi.org/10.64898/2026.02.08.704659doi: bioRxiv preprint 4 replace the extracellular domain with a single-chain variable fragment or nanobody that binds a user-defined ligand (e.g. GFP), and the intracellular domain with a transcriptional activator that drives a chosen output gene (Morsut et al. 2016). Upon binding to a ligand on a contacting cell, synNotch is cleaved by intramembrane proteolysis by endogenous factors, releasing the transcriptional activator to enter the nucleus and activate transgene expression. SynNotch has been widely used in vitro to report cell–cell contacts and program cellular behaviors (Morsut et al. 2016; Roybal et al. 2016; Toda et al. 2018; Choe et al. 2021; Malaguti et al. 2022; Reddy et al. 2024). It has been extended in vivo to trace contacts during mouse vascular development (Zhang et al. 2022). In Drosophila, this system has been adapted as TRACT, demonstrating that synNotch-based contact can detect anterograde synaptic contacts in certain circuits (Huang et al. 2016; Huang et al. 2017). These advances suggest that synNotch is a promising platform for genetically encoded transsynaptic tracing in mammals. Here, we adapt synNotch in mice to create an anterograde, AAV-deliverable tracing system called TRACR - Transneuronal Anterograde Circuit Readout. Through a series of experiments in the mouse visual system, we show that TRACR: 1) can be delivered using non-toxic AAV vectors; 2) can be restricted to genetically defined starter cells; 3) effects efficient and specific labeling of functional postsynaptic partners across both long-range and local connections; and 4) reverses labeling of postsynaptic neurons as result of degeneration of input neurons. TRACR provides a modular, accessible transneuronal tracing system that is ready to use to study any circuit of interest in the mouse central nervous system, as well circuit remodeling in development and disease. .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 8, 2026. ; https://doi.org/10.64898/2026.02.08.704659doi: bioRxiv preprint 5

TRACR system design To develop a ligand-receptor based tracing method for mammalian circuits, we adapted the synthetic Notch (synNotch) system to convert synaptic contact into a reporter signal that can be visualized in vivo. Building on canonical synNotch designs (Morsut et al. 2016), we designed a three-component Sender-Receiver-Reporter system that signals as follows (Figure 1A): ‘Sender’ cells display a membrane-tethered GFP ligand at presynaptic terminals. ‘Receiver’ neurons express a chimeric synNotch receptor comprising an extracellular anti-GFP nanobody (LaG17), the Notch core regulatory and transmembrane domains, and a cleavable intracellular transcriptional module, tetracycline-controlled transactivator (tTA). Ligand binding exerts mechanical force across the Notch core, triggering the endogenous proteolytic cascade, and liberating the tTA to translocate to the nucleus and drive expression of a Reporter transgene under tTA-responsive elements (TRE). We designed the components to be delivered via adeno- associated virus (AAV), targeting Sender expression to genetically defined presynaptic populations using Cre/LoxP strategies or cell type-specific promoters, and enabling identification of target cells among ubiquitous Receiver and Reporter-expressing neurons (Figure 1B). A critical challenge in adapting synNotch for transneuronal tracing is to confine contact- dependent signaling to synaptic interfaces. After evaluating previous synapse-targeting strategies (Kim et al. 2011; Yamagata and Sanes 2012; Martell et al. 2016; Coomer et al. 2025), we tethered the GFP ligand to the transmembrane and cytoplasmic domains of Neurexin-3β (Figure 1B). These Neurexin-3β domains are highly similar to those in Neurexin-1β, and promote targeting to axon terminals and surface membrane insertion (Fairless et al. 2008; Gokce and Südhof 2013; Klatt et al. 2021). To minimize potential artifacts, we omitted the extracellular domains of Neurexin-3β which can drive ectopic synaptogenic interactions when overexpressed (Kim et al. 2011; Tsetsenis et al. 2014). To augment cellular processing of the Nrx3β (NRX)-GFP ligand, we incorporated signal peptide and membrane trafficking sequences shown to improve surface delivery of engineered opsin proteins (Zhao et al. 2008; Gradinaru et al. 2010). The Sender AAV constructs encoding the NRX-GFP ligand include a promoter of choice or a Cre-dependent design to restrict expression to genetically identified presynaptic neuronal populations (Figure 1B). A modular bicistronic design enables co-expression of NRX–GFP with .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 8, 2026. ; https://doi.org/10.64898/2026.02.08.704659doi: bioRxiv preprint 6 fluorescent markers, recombinases or effectors to support downstream investigations such as anatomical tracing and optogenetic interrogations (Figure 1B). In cultured neurons, co- transfection of an AAV Sender (AAV-DIO-NRXGFP-T2A-FlpO) along with AAV-Cre produced robust expression and trafficking of NRX–GFP throughout neurites (Figure 1C). Consistent with surface presentation, extracellular GFP staining under non-permeabilized conditions labeled NRX–GFP puncta that overlap with a presynaptic marker, synapsin-1 (Figure 1C). For unbiased tracing of downstream postsynaptic cells, we created Receiver-AAV plasmids to express synNotch ubiquitously in neuronal cells without post-synaptic tethering (Figure 1B). This choice was intended to minimize interference with synNotch activation, and to permit detection across diverse excitatory and inhibitory synapse types, which share few proteins at the postsynaptic density, including neuroligins (Bemben et al. 2015). The tTA (Tet-OFF) was selected as the transcriptional effector based on its demonstrated performance in synNotch assays and the extensive implementation of tTA-TRE transgenic platforms in the mouse nervous system (Madisen et al. 2015; Morsut et al. 2016; Malaguti et al. 2022; Zhang et al. 2022). Transfection of Receiver-AAV in cultured neurons confirmed robust surface expression of the Myc-tagged LaG17-synNotch-tTA fusion proteins along neuronal processes (Figure 1D). For tTA-dependent reporter expression, we used available AAVs (Chan et al. 2017) and transgenic mice expressing red fluorescent protein (RFP) from the TRE promoter (Ai63-TIT-tdT, (Daigle et al. 2018)) (Figure 1B). To minimize the ligand-independent activation reported in contexts of high synNotch expression, we constructed a second Receiver with a hydrophobic protein sequence inserted in the C-terminus of the Notch core (Figure 1B, Receiver-RAM7 (Yang et al. 2020)). We designated this entire system as TRACR – TRansneuronal Anterograde Circuit Readout.We posit that in TRACR, NRX–GFP on presynaptic terminals binds synNotch on postsynaptic partners to release tTA and drive Reporter expression, resulting RFP labeling of synaptic targets. We test the specificity and sensitivity of TRACR signaling across multiple synapses in the mouse visual system. TRACR identifies long-range connections We first sought to assess the TRACR system in vivo in the retinothalamic projections. The mouse retinothalamic projection has well-defined anatomical boundaries, where retinal ganglion .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 8, 2026. ; https://doi.org/10.64898/2026.02.08.704659doi: bioRxiv preprint 7 cells (RGCs) send axons to the thalamus and form dense synaptic terminations almost exclusively within the dorsal lateral geniculate nucleus (dLGN). We predicted that TRACR signaling will induce reporter expression in dLGN neurons, but not in the surrounding thalamic nuclei that lack RGC inputs, such as the ventral posteromedial nucleus (VPM) (Martersteck et al. 2017). To visualize the NRX-GFP ligand within RGC terminals in the dLGN, we co-injected Cre- dependent AAVs encoding the Sender and an HA-tagged presynaptic marker (Chantranupong et al. 2020) into retinas of vGluT2-Cre mice, which label almost all RGCs (Martersteck et al. 2017) (Figure 1E). GFP positive axon terminals were distributed throughout known retinal projection fields and overlapped with the HA-synaptophysin, confirming a presynaptic localisation of the Sender-GFP ligand (Figure 1F-G). We next delivered the AAVs encoding the Receiver and TRE-mRuby Reporter via stereotaxic injection to the thalamus, encompassing the dLGN and surrounding nuclei. The myc-tagged Receiver was broadly expressed across the injected thalamic area, including the dLGN, the ventral lateral geniculate nucleus (vLGN), VPM, posterior complex, and lateral dorsal nucleus (Figure 1H-I). To control for ligand-independent activation and reporter leakiness, we injected Receiver and Reporter AAVs at different dilutions, and without Sender virus. While delivery of high titer of Reporter AAV caused leaky RFP signals, delivery of increasingly diluted AAVs revealed graded decrease in reporter signals to absent or rare RFP-positive cells (Supp. Figure 1). Thus, titrating dosages of the Receiver and Reporter AAVs in the absence of Sender AAV in the region of interest is required for non-leaky tracing, similar to titrating helper and reporter AAVs for retrograde tracing using rabies virus (Lavin et al. 2020). When all TRACR components were delivered, RFP expression induced by synNotch signaling in putative postsynaptic partners was almost exclusively restricted to the dLGN (Figure 1I). The dense reporter labeling within the dLGN, coupled with its conspicuous absence in the adjacent Receiver-expressing but non-synaptically-connected nuclei, demonstrates that TRACR effectively identifies targets within a long-range anatomical projection. TRACR identifies interneuron targets in local circuitry We next tested TRACR to dissect local neuronal connections, a task that remains challenging with established transsynaptic tracers. To evaluate TRACR’s resolution and specificity in a dense short-range network, we turned to the inner plexiform layer (IPL) of the retina (Figure 2A). .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 8, 2026. ; https://doi.org/10.64898/2026.02.08.704659doi: bioRxiv preprint 8 The IPL is an ideal test site where neurites of over 100 molecularly defined neuron types stratify into specific sublaminae to establish distinct functional circuits (Shekhar et al. 2016; Tran et al. 2019; Yan et al. 2020). Focusing on a well-known circuit for direction selectivity, we tested the specificity of TRACR for tracing a local circuit by evaluating the: i) morphology of Reporter- positive neurons, including overlap with pre-synaptic neurites, ii) expression of appropriate molecular markers, and iii) functional analysis. We evaluated TRACR in retinas of ChAT-Cre mice, which express Cre exclusively in starburst amacrine cells (SACs) (Lefebvre et al. 2012) (Figure 2B,C). SACs are GABAergic interneurons that shape the movement responses of direction-selective retinal ganglion cells (DSGCs). Their dendrites stratify in two IPL sublaminae (s2 and s4), where they synapse primarily onto bistratified ON–OFF DSGCs and onto other SACs (Yoshida et al. 2001; Briggman et al. 2011; Yonehara et al. 2011; Chen et al. 2016; Ding et al. 2016; Brombas et al. 2017; Mauss et al. 2017), with additional RGC types receiving weaker innervation (Beier et al. 2013; Baden et al. 2016; Wang and Zhang 2023). We predicted that TRACR Reporter activation will occur mainly in SACs and DSGCs, with Sender and RFP-positive neurites co-stratifying in s2 and s4 (Figure 2C). TRACR Reporter induction required all three components: co-injection of Receiver and Reporter AAVs without Sender, or Sender and Reporter AAVs without Receiver, yielded no detectable RFP at P60 (Figure 2B). When all TRACR AAVs are delivered in ChAT-Cre retinas, Reporter- positive neurons were observed in both the inner nuclear layer (INL) and ganglion cell layer (GCL), with processes overlapping with Sender neurites in s2 and s4 in the IPL (cosine similarity index = 0.89± 0.05, Figure 2D,E). To compare the SAC-driven TRACR output to other presynaptic neurons, we performed parallel experiments in Gad2-Cre mice, which label a broader class of GABAergic amacrine cells (Figure 2F). As expected, Sender neurite lamination in Gad2-Cre retinas was broader than observed in Chat-Cre retinas, but these neurites still overlapped Reporter labelled neurites (cosine similarity index, 0.93 ± 0.07; n= 4 retinas from 4 animals; Figure 2G,H). Importantly, reporter signals in Gad2-Cre retinas were significantly different than that of Chat-Cre (cosine similarity index = 0.60 ± 0.02, p<0.01 vs Chat-Cre and p=0.01 vs Gad2-Cre), confirming that TRACR reporter signal is dependent on the Sender population. Molecular characterization of TRACR signals supported circuit specificity: 89.6% of reporter- positive neurons in ChAT-Cre retinas co-expressed either ChAT (SACs) or SATB1 (ON–OFF .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 8, 2026. ; https://doi.org/10.64898/2026.02.08.704659doi: bioRxiv preprint 9 DSGCs), markers for SAC target cells (189/211 cells, n = 5 retinas) (Lefebvre et al. 2012; Peng et al. 2017) (Figure 2I). In contrast, only 27.8% (33/209 cells, n = 3 retinas) of reporter-positive neurons in Gad2-Cre retinas expressed markers of SACs or ooDSGCs (Figure 2J). Consistent with known wiring, no reporter co-localization was detected (0/126 and 0/101 cells, respectively; n = 3 retinas) with SMI-32 (α-RGCs) or Opn4 (ipRGCs) in Chat-Cre retinas, two RGC types not receiving SAC input (Viney et al. 2007; Baden et al. 2016) (Figure 2I,K). As expected, Gad2-Cre TRACR labeling included these cell types at frequencies consistent with their population abundance (α-RGCs, 5.1%; ipRGCs, 6.0%) (Hattar et al. 2002; Tran et al. 2019) (Figure 2J,K). We next performed loose-patch recordings from Reporter-positive and -negative RGCs in ChAT-Cre retinal explants treated with TRACR (Figure 3A,B). RFP-positive RGCs showed predominately ON-OFF responses to a full field flash, however some ON- and OFF-RGCs were encountered (Figure 3C). This was in contrast with random sampling of nearby RFP-negative RGCs which tended to be predominantly ON- or OFF- RGCs. These results are consistent with the expected frequency of ON, OFF, and ON-OFF RGC responses to full field stimuli and suggested that RFP+ RGCs may be enriched for ON-OFF DSGCs. To examine this idea, we analyzed RGC responses to bright bars passed over their receptive field center moving in 8 different directions. As expected, many ON-OFF RFP-positive RGCs fired brief bursts of spikes when the bright bar’s leading edge entered the receptive field (ON) and again when the bar’s trailing edge exited the receptive field (OFF, Figure 3D). Polar plots of firing rate from these RFP-positive ON-OFF RGCs were often asymmetric indicating that these neurons fired most to bars movement in a particular direction (Figure 3E). Average ON- and OFF- direction selective indices computed from all RFP-positive and RFP-negative RGCs confirmed this picture and showed significantly higher average DSIs for RFP-labelled neurons (Figure 3F). Finally, we plotted angular preferences aligned to the cardinal poles of retina for only ON-OFF RFP-positive RGCs and saw biased tuning towards dorsal, ventral, nasal and temporal directions (Figure 3G). These results support our anatomical experiments and show that TRACR outputs from SACs predominantly labels ON-OFF direction selective RGCs. Altogether, these data provide anatomical, molecular and functional evidence that TRACR labels the postsynaptic targets of genetically defined retinal neurons. They also highlight TRACR’s selectivity and utility for tracing local connections, including those of interneurons, where Senders and Receivers are expressed by physically proximate neurons. .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 8, 2026. ; https://doi.org/10.64898/2026.02.08.704659doi: bioRxiv preprint 10 TRACR signals specifically across photoreceptor–bipolar cell synapses To evaluate the sensitivity and specificity of TRACR signalling, we tested the system in the photoreceptor–ON bipolar cell circuit, a well-characterized synapse in which all contacts are confined to the outer plexiform layer (OPL), making them straightforward to visualize. We created a Sender-AAV with a photoreceptor-specific promoter (ProC1, (Jüttner et al. 2019)) to target NRX-GFP expression in both rod and cone photoreceptors, and a Receiver-AAV with an ON bipolar-specific promoter to drive expression in ON cone and rod bipolar cells (4xGrm6 promoter (Lagali et al. 2008)) (Figure 4A). We injected AAVs in retinas of P5 Ai63-TRE- TdTomato (Ai63-TdTomato) reporter animals and used the Receiver with the modified synNotch-RAM7 to minimize ligand-independent activation observed in initial experiments. By restricting Sender and Receiver expression to known synaptic partners, we explicitly evaluated whether TRACR reports postsynaptic partner identity and synaptic connectivity, rather than proximity alone, and whether signaling is reversible and dependent on functional synapses. A key requirement for TRACR signalling is that the ligand, NRX-GFP, localizes to the presynaptic terminals of Sender-expressing cells. Unamplified GFP fluorescence showed concentrated ligand at photoreceptor terminals in the OPL, and closely colocalized with presynaptic proteins Ribeye and PSD95 (Figure 4B). In subsequent experiments, anti-GFP staining identified Sender-positive photoreceptors but showed a broader distribution within the cells consistent with antibody detection of low-level intracellular protein (Figure 4C). Together, these observations confirm ligand enrichment at presynaptic sites, satisfying a prerequisite for TRACR activation. We next verified that Grm6-driven Receiver constructs were expressed in bipolar cells. Immunostaining at P20 showed that the Myc-tagged Receiver was confined to cells in the apical region of the INL displaying bipolar-like morphologies, consistent with Grm6 promoter activity in bipolar cells (Figure 4C). Approximately 83% of Receiver-positive cells were immunoreactive for Gαo, a pan–ON bipolar cell marker (Supp. Figure 2). The remaining ~11% of Gαo negative Receiver-positive cells expressed Chx10, a pan-bipolar cell marker (Supp. Figure 2). Thus, Receiver expression is largely restricted to ON bipolar cells and nearly all Receiver-positive cells belong to the bipolar cell lineage. These findings demonstrate selective and efficient expression of TRACR components within the photoreceptor–bipolar cell circuit. .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 8, 2026. ; https://doi.org/10.64898/2026.02.08.704659doi: bioRxiv preprint 11 To determine whether TRACR activation depends on Sender-Receiver interaction, we compared reporter signal at P20 in Ai63-TdTomato mice injected with AAV Receiver alone versus those co-injected with AAV Sender and Receiver. Receiver-only injections produced minimal reporter signal in bipolar cells, indicating low ligand-independent activation (Figure 4C). In contrast, co-injection of Sender and Receiver produced robust RFP reporter activation that was restricted to Receiver-expressing bipolar cells. In contrast, Sender+Receiver co-injection induced robust RFP expression restricted to Receiver-positive bipolar cells, significantly increasing the proportion of Reporter-positive cells within the Receiver-positive population compared with eyes injected with Receiver alone (Figure 4C). These data demonstrate that TRACR activation is confined to Receiver-expressing cells and is dependent on the presynaptic ligand. Finally, reporter-positive cells co-expressed either PKC, a marker of rod bipolar cells, or SCGN, a marker of cone bipolar cells (Figure 4D). Quantitative analysis confirmed significant TRACR activation in rod and cone bipolar populations, and that nearly all reporter-positive cells were bipolar cells (Figure 4D). Therefore, TRACR detects synaptic connectivity across both rod and cone pathways. To determine whether TRACR detects established synapses in the mature retina, we delivered the Sender AAV subretinally at postnatal day 45 (P45) into eyes that had received the Receiver intravitreally at P5, as efficient bipolar-driven Receiver expression could not be achieved following adult delivery. After fifteen days, reporter activation was again observed in Receiver- positive bipolar cells, and at levels comparable to those following neonatal virus delivery (Supp. Figure 3B). Ligand-independent reporter activity in Receiver-only controls remained similarly low (Supp. Figure 3). Although Receiver expression was also observed in amacrine cells located in the basal INL at these timepoints, these cells did not activate the reporter, further confirming that direct Sender inputs are required for TRACR activation (Supp. Figure 3). Together, these experiments demonstrate that TRACR effectively labels postsynaptic partners in both developing and adult retinal circuits. Our next questions focused on validating that TRACR activation reflects trans-synaptic interactions, rather than proximity between the Sender ligand and the Receiver receptor. To test this, we used two complementary strategies. First, we expressed both Sender and Receiver in photoreceptors (Figure 4E). If TRACR is activated within the same cells or between adjacent cells, this condition should generate reporter signal. However, we did not detect any reporter expression (Figure 4E). Second, we expressed the Sender in Müller glia and the Receiver in bipolar cells. We generated a Müller glia–specific AAV-Sender (using ProB2 enhancer, (Jüttner .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 8, 2026. ; https://doi.org/10.64898/2026.02.08.704659doi: bioRxiv preprint 12 et al. 2019)) and confirmed its expression in Müller glia by co-immunostaining with Sox9 (Supp. Figure 4). Müller glial processes ensheath photoreceptor synapses (Burris et al. 2002; Williams et al. 2010), placing the Sender ligand in proximity to Receiver-expressing bipolar dendrites. Despite this anatomical proximity, reporter activation was absent or limited to a few cells (Figure 4E). Together, these experiments demonstrate that TRACR is not activated by mere spatial proximity, but instead requires presynaptic Senders in photoreceptors and postsynaptic Receivers in bipolar cells, establishing its specificity for trans-synaptic contacts. TRACR reports loss of photoreceptor–bipolar cell connectivity following photoreceptor degeneration Most tracing tools rely on recombinase-driven reporters that permanently mark synaptic partners, so that labels persist even after synapses are eliminated or remodeled. TRACR instead uses the reversible tTA–Tet system, which should require ongoing synNotch cleavage and transcription driven by intact synaptic contact. We therefore asked whether TRACR activation is lost upon loss of photoreceptor–bipolar cell connectivity, a fundamental requirement for using this system to evaluate disease models or genetic perturbations that disrupt synapses. We tested TRACR activation in two photoreceptor degeneration models. We first used the rapid genetic rod degeneration model, rd1 (Bowes et al. 1990), and delivered the photoreceptor- specific Sender and bipolar-specific Receiver AAVs in retinas of P5 rd1; Ai63 animals (Figure 5A). Prior to substantial photoreceptor loss, Sender and Receiver expression as well as TRACR Reporter activation were robust, indicating that compromised photoreceptors in this model can still express the constructs and induce TRACR signaling across synaptic contacts (Figure 5B). By P30, when most rods have degenerated, reporter expression was significantly reduced, and by 2 months it was nearly undetectable (Figure 5B). Because rd1 is an early-onset degeneration model, we also assessed acute photoreceptor loss in adult retinas, after establishment of photoreceptor-bipolar synapses. We induced degeneration with MNU, which rapidly eliminates photoreceptors (Petrin et al. 2003). In MNU-treated eyes, Sender-positive cells were lost and TRACR activation was significantly reduced compared to controls (Figure 5C). Thus, TRACR signal declines as synaptic contacts are eliminated, and the readout depends on the continued presence of presynaptic photoreceptors. Together, these results show that TRACR reversibility enables analysis of synapse disruption in disease and injury models. .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 8, 2026. ; https://doi.org/10.64898/2026.02.08.704659doi: bioRxiv preprint 13 TRACR detects synaptic disruption in genetic and activity-dependent models We then investigated whether TRACR can detect synaptic dysfunction, as this would enhance its utility for assessing how disease models or genetic modifications impact connectivity. To do this, we used two approaches. First, we blocked synaptic vesicle release in rods by conditionally expressing a Tetanus toxin gene in Nrl-Cre; R26LSL-TeNT mice (Zhang et al. 2008) (Figure 6A). This intervention abolishes SNARE-mediated synaptic vesicle release and disrupts synapse formation between rods and bipolar cells (Cao et al. 2015). In this instance, TRACR activation in rod bipolar cells was significantly decreased in Rod-TeNT mice compared to controls (Figure 6B,C). Second, we evaluated TRACR activation in Nrl-Cre; Elfn1fl/fl mice with a rod photoreceptor-specific inactivation of Elfn1, a protein essential for the establishment of rod-to-rod bipolar cell synapses (Cao et al. 2015). In this instance, reporter induction was significantly reduced in Rod-Elfn1 cKO mice compared to controls (Figure 6D,E). Interestingly, in both the Rod-TeNT and the Rod-Elfn1 cKO mice, most of the remaining RFP-positive cells were SCGN-positive cone bipolar cells, consistent with preserved cone-to–cone bipolar connectivity when rod pathways are selectively impaired (Figure 6C,E). Together, these findings show that in two independent models of rod dysfunction, TRACR selectively reports intact cone connectivity. In summary, TRACR detects functional rod-to–rod bipolar cell synapses and fails to activate when synapses are lost, silent, or absent, demonstrating that the system provides a reversible readout of synapse loss and reports genuine alterations in photoreceptor–bipolar connectivity. .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 8, 2026. ; https://doi.org/10.64898/2026.02.08.704659doi: bioRxiv preprint 14

A persistent limitation in circuit mapping is the lack of broadly adoptable anterograde tracing tools that also provide separable genetic access to each side of a connection. We developed TRACR, an AAV-deliverable toolkit that adapts synthetic Notch signaling to convert presynaptic ligand binding into tTA-TRE driven reporter expression in postsynaptic cells. We showed that, across multiple synapses in the mouse visual system, TRACR labeled targets downstream of sensory neurons, inhibitory interneurons, and long-range projection neurons, and distinguished partners within local circuits. Because TRACR signaling requires an intact synapse, its transcriptional output provides a temporally sensitive readout of connection status, including changes resulting from degeneration. Since each element is modular and compact enough to fit into typical AAV vectors, TRACR offers immense potential for transneuronal control of gene expression that can be easily integrated with existing methods for genetic markers, sensors, or optogenetic actuators. Together, these results establish TRACR as a broadly deployable platform for anterograde transneuronal tracing and longitudinal circuit analysis in mammals. Tracing advances offered by TRACR Because TRACR does not depend on virus or protein crossing the synaptic cleft, it is unique among mouse transneuronal tracers in using a defined synNotch-based ligand–receptor interaction that converts presynaptic contact into a postsynaptic transcriptional readout. This is in contrast to viral anterograde approaches, whose transfer mechanisms are ill-defined with potential for bidirectional or polysynaptic spread, and whose efficiency varies with cell type and circuit context (Beier 2019; Xu et al. 2020). Benchmarking TRACR at the photoreceptor–bipolar synapse showed that induction is Sender-dependent, with minimal ligand-independent activation, and synapse-dependent, as signaling ceases when synapses are disrupted. TRACR’s anterograde directionality is determined by presynaptic enrichment of the GFP ligand, while its labeling is inherently monosynaptic because it produces postsynaptic transcription, rather than a transmissible virus or protein. Unlike many viral approaches prone to cellular toxicity, TRACR was non-toxic and did not measurably perturb synapse integrity or circuit function. We showed this in three ways. . First, Sender and Receiver expression in multiple retinal cell types did not cause overt cell loss or retinal thinning. Second, TRACR did not impair synapse formation or integrity. TRACR signaling persisted in wild-type retinas over many weeks, whereas signaling declined or was absent in retinas with photoreceptor degeneration (rd1, MNU treatment), neurotransmission blockade (TeNT), or impaired synapse formation (Elfn1 cKO). Third, TRACR-positive retinal ganglion cells .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 8, 2026. ; https://doi.org/10.64898/2026.02.08.704659doi: bioRxiv preprint 15 retained normal ON–OFF responses and direction tuning. Together with its reliance on standard AAV delivery, these features make TRACR broadly accessible compared to pathogenic tracers that have greater biosafety and regulatory constraints. An additional strength is TRACR’s modularity and genetic access to either side of the synapse. Sender expression can be restricted to presynaptic populations using existing Cre lines and the expanding repertoire of AAV enhancers and promoters (Hrvatin et al. 2019; Jüttner et al. 2019; Graybuck et al. 2021; Ben-Simon et al. 2025; Furlanis et al. 2025 Apr 22). This contrasts to approaches such as AAV1-Cre or YFV-17D that transfer Cre itself and limit presynaptic specificity (Zingg et al. 2017; Zingg et al. 2020; Li et al. 2021). For unbiased partner discovery, Receiver and Reporter can be expressed broadly so that postsynaptic labeling is agnostic to target identity, provided that AAV infection is uniform. TRACR can leverage tTA-dependent AAVs and transgenic mice to mark post-synaptic populations with reporters of choice, including sensors, or optogenetic actuators (Daigle et al. 2018). The TRACR design is interchangeable with common Cre, Flp and other transcriptional systems, enabling an almost infinitely customisable set of markers, effectors, and sensors, with the main practical constraint being AAV packaging capacity. Finally, TRACR appears effective across neuron classes and synapse types, and it resolves local circuits, which remains difficult for most transsynaptic tools. In the visual system, TRACR labeled predicted postsynaptic partners of sensory neurons (photoreceptors), GABAergic interneurons (amacrine cells), and projection neurons to the brain (retinal ganglion cells), indicating that the synNotch-tTA induction can report a variety of connections. The retina’s inner plexiform layer (IPL) is an especially stringent setting for testing TRACR in local wiring. The IPL is a dense in which processes from over 100 bipolar, amacrine, and ganglion cell types intermingle yet form stereotyped microcircuits. In this context, TRACR distinguished targets of starburst amacrine cells from those of broader amacrine populations. Proximity and co- stratification of processes are insufficient to infer connectivity in the IPL (Helmstaedter et al. 2013; Bae et al. 2018; Prigge et al. 2023), with clear examples of synapse partnering choices determined by recognition molecules and required for circuit output (Krishnaswamy et al. 2015; Rochon et al. 2021; Friedrichsen et al. 2024; Olguin et al. 2025). As the connectivity of many retinal circuits remains unresolved, TRACR stands to be an ideal tool for mapping synaptic partnerships. .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 8, 2026. ; https://doi.org/10.64898/2026.02.08.704659doi: bioRxiv preprint 16 TRACR as a mammalian ligand–receptor tracing system. TRACR represents the first mammalian implementation of a ligand–receptor tracing strategy that couples synaptic contact to transcriptional output. Conceptually, it parallels TRACT and Trans-Tango in Drosophila, which similarly convert synaptic contact into gene expression using synNotch or engineered GPCR signalling, respectively (Huang et al. 2017; Talay et al. 2017). A key distinction is that synNotch-based systems rely on endogenous proteases such as γ- secretase, whereas Trans-Tango uses reconstituted TEV protease recruited to an engineered receptor. In zebrafish, Trans-Tango has been combined with optogenetic and calcium imaging tools to probe functional coupling in vivo (Coomer et al. 2025). Specificity, sensitivity, and limitations of TRACR Across experiments, we found that viral dose, expression levels and basal activation of Receiver/Reporter AAV components can affect TRACR’s performance. Consistent with prior reports of Tet-off leakiness and ligand-independent synNotch activation, we observed low-level reporter expression with Reporter alone or Receiver + Reporter in the absence of Sender In our hands, empirical titration of Receiver and Reporter AAVs minimized this background. Thus, inclusion of appropriate controls is an important consideration for experimental designs using the AAV complement of TRACR. Although tTA-independent activation was not detected in Ai63-TITL-tdTomato mice, we observed ligand-independent activation from the original Receiver using this line. Reduced reporter expression to minimal or undetectable levels in Ai63 controls and improved specificity was achieved by replacing Receiver with synNotch-RAM7, which contains a hydrophobic sequence native to the Notch core that suppresses nonspecific activation (Yang et al. 2020). With synNotch-RAM7, TRACR achieved high sensitivity at the photoreceptor–bipolar synapse (~60% induction among Receiver-positive cells in wildtype retinas), with variability likely reflecting infection efficiency. Sensitivity was lower in some contexts, such as Chat-Cre driven starburst amacrine outputs. Increasing Sender expression with strong promoters, higher Sender:Receiver ratios (Malaguti et al. 2022), and allowing longer infection times for long-range axons (4–6 weeks) can improve induction. TRACR components should be independently titrated for each circuit, in part because AAV tropism and surgical accessibility constrain expression levels. Another limitation is that the GFP ligand uses neurexin-3beta intracellular domains that promote surface delivery and presynaptic enrichment (Fairless et al. 2008; Gokce and Südhof 2013; Klatt .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 8, 2026. ; https://doi.org/10.64898/2026.02.08.704659doi: bioRxiv preprint 17 et al. 2021). This aligns with the strong presynaptic enrichment of the GFP ligand observed at photoreceptor and RGC terminals. We posit this enrichment increases TRACR signaling over threshold reporter induction to connected partners, rather than in processes that may come into contact. Consistent with this idea, Sender-expressing photoreceptors did not drive reporter induction in adjacent photoreceptors or Müller glia. However, stable neurexin complex formation at synapses typically involves extracellular interactions with neuroligins, but these domains can risk ectopic synaptogenesis when overexpressed (Kim et al. 2011; Tsetsenis et al. 2014). Additional strategies to optimize the pre- and postsynaptic restriction of TRACR components are being explored. As with any transsynaptic approach, TRACR output should be interpreted with the possibility of false positives (background activation or signaling at closely apposed neurites) and false negatives (missed partners due to limited sensitivity). Accordingly, TRACR experiments should be validated using complementary anatomical, electrophysiological, or optogenetic assays. In summary, TRACR offers a unique, reversible transneuronal tracing method for unbiased, high-throughput identification of postsynaptic partners and longitudinal analysis of circuit connectivity in mammalian systems. By coupling synaptic contact to transcription, TRACR can track connectivity gains and losses over time in studies of disease models or genetic perturbations that disrupt synapses. .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 8, 2026. ; https://doi.org/10.64898/2026.02.08.704659doi: bioRxiv preprint 18

Animal care and mouse lines Animal studies were carried out in accordance with the guidelines of the Canadian Council on Animal Care, and under protocols approved by the Animal Care Committees of the Centre for Phenogenomics (TCP), Laboratory Animal Services (LAS) at the Hospital for Sick Children, University Health Network Research Institute (Toronto Canada) and the Goodman Cancer Research Facility Vivarium at McGill University (Montreal, Canada). All facilities are certified by the Canadian Council on Animal Care and while TCP, LAS, and University Health Network facilities registered under the Animals for Research Act of Ontario. Mice were maintained on a C57/B6J background. Both male and female mice were used at all ages. The following mouse lines were purchased from the Jackson Laboratory: ChAT-ires-Cre (B6;129S6-Chattm2(cre)Lowl/J, Strain #:006410; RRID:IMSR_JAX:006410) (Rossi et al. 2011); Gad2-ires-Cre (Gad2tm2(cre)Zjh/J; Strain #:010802; RRID:IMSR_JAX:010802) (Taniguchi et al. 2011); Vglut2-ires-Cre (B6J.129S6(FVB)-Slc17a6tm2(cre)Lowl/MwarJ, Strain #:028863; RRID:IMSR_JAX:028863) (Vong et al. 2011); Nrl-Cre (C57BL/6J-Tg(Nrl-cre)1Smgc/J, Strain #:028941; RRID:IMSR_JAX:028941) (Brightman et al. 2016); and Rd1 mice with retinal degeneration phenotype (C57BL/6J-Pde6brd1-2J/J (Strain #:004766; RRID:IMSR_JAX:004766). Ai63(TIT-tdTomato) are tTA-dependent reporter mice expressing tdTomato from the TRE-tight promoter, and were created by targeted insertion at the Igs7 TIGRE locus (B6.Cg-Igs7tm62.2(tetO- tdTomato)Hze/J)(Daigle et al. 2018). Ai63 mice were obtained from the Allen Institute for Brain Science (a kind gift from Dr. Hongkui Zeng, Seattle, USA). R26LSL-TeNT (Gt(ROSA)26Sortm2(GFP/tetX)Gld) (Zhang et al. 2008) animals were a kind gift from Dr. Michel Cayouette (Institut de Recherches Cliniques de Montréal, Montreal, Canada). The Elfn1fl/fl were described previously (Cao et al. 2015). Plasmid cloning Sender plasmids were generated as follows: To create the NRX-GFP ligand, codon-optimised sequences of the transmembrane and cytoplasmic domains of the human Neurexin-3β (NP_620426.2) followed by the membrane trafficking signal of human Kir2.1 (NP_000882.1) (Gradinaru et al. 2010) were synthesized as gBlocks (Integrated DNA Technologies). eGFP was amplified from pCAGIG (Matsuda and Cepko 2004) with the addition of the N-terminal signal peptide sequence of rat β2 acetylcholine receptor (Zhao et al. 2008). To create pAAV.hSyn- DIO-Nrx3b-GFP-T2A-smFPHA-WPRE-pA, fragments were fused using Gibson assembly (# .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 8, 2026. ; https://doi.org/10.64898/2026.02.08.704659doi: bioRxiv preprint 19 E2611S, New England Biolabs) into a pCAG-driven backbone to create a transient vector. Nrx3b-GFP and smFPHA derived from pCAG-smFP_HA (gift from Loren Looger; Addgene plasmid # 59759 ; http://n2t.net/addgene:59759 ; RRID:Addgene_59759) (Viswanathan et al. 2015) were PCR amplified with overlapping tails to insert a T2A sequence, and assembled between the loxP sites of a linearized hSyn-containing AAV backbone by Gibson assembly (pAAV.hSyn-DIO-hM3D(Gq)-mCherry, a gift from Bryan Roth, Addgene plasmid # 44361; http://n2t.net/addgene:44361 ; RRID:Addgene_44361) (Krashes et al. 2011). To create a Sender with a channelrhodopsin marker (pAAV.hSyn-DIO-Nrx3b-GFP-T2A-hChR2(H134R)- eYFP-WPRE-pA), the smFPHA insert was replaced with hChR2(H134R)-eYFP (Addgene plasmid #26973, a gift from Karl Deisseroth; http://n2t.net/addgene:26973; RRID:Addgene_26973). To create CBh-driven Senders (pAAV.CBh-DIO-Nrx3beGFP-T2A- ChR2(H132R)-eYFP-WPRE3-pA), the Sender insert was subcloned into AAV backbone (AiP11839-pAAV.hSyn1-SYFP2-10aa-H2B-WPRE3-BGHpA, Addgene plasmid # 163509; http://n2t.net/addgene:163509; RRID:Addgene_163509a; gift from Allen Institute for Brain Science & Jonathan Ting) (Graybuck et al. 2021) and the hSyn promoter was replaced with CBh (from pAAV.CBh-DIO-LifeAct-eGFP-WPRE) (Ing-Esteves and Lefebvre 2024). Sender plasmid with T2A-Flpo (pAAV.hSyn-DIO-Nrx3beGFP-T2A-Flpo-WPRE) allows expression of Flp- dependent vectors to complement Sender expression. pAAV.hSyn-DIO-Nrx3beGFP-T2A-Flpo was created by amplifying a FlpO sequence from Addgene plasmid #26745 (gift from Rolf Zeller; http://n2t.net/addgene:26745; RRID:Addgene_26745) (Osterwalder et al. 2010) which was then inserted into pAAV.DIO-Sender-GFP backbone by Gibson Assembly. To create retina cell-type-specific Senders, synthetic promoters developed by Jüttner et al. (Jüttner et al. 2019) were PCR amplified or subcloned from the following plasmids: pan- photoreceptor ProC1 (pAAV.ProC1-CatCh-GFP-WPRE, gift from Botond Roska, Addgene plasmid # 12593 ; http://n2t.net/addgene:125937 ; RRID:Addgene_125937); Müller Glia ProB2 (pAAV.ProB2-CatCh-GFP-WPRE, Addgene plasmid # 125922; http://n2t.net/addgene:125922 ; RRID:Addgene_125922). Amplified fragments were inserted into a linearized Sender-AAV backbone using standard restriction digest and ligation. The ChR2 marker was replaced by Jaws-KGC-ERT2 optogenetic channel (Chuong et al. 2014) tagged with hemagglutinin (HA). The Jaws-KGC-ERT2-HA insert was synthesized (Bio Basic Inc., Markham, Canada) and inserted into the linearized backbone to create pAAV.ProX-Nrx3b-GFP-T2A-Jaws-HA-WPRE- pA. .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 8, 2026. ; https://doi.org/10.64898/2026.02.08.704659doi: bioRxiv preprint 20 The Receiver construct (pAAV.hSyn-LaG17-SynNotch-tTA-WPRE-pA) was created by PCR amplification of sequences encoding the LaG17 nanobody, SynNotch and TetROFF from pHR- SFFV-LaG17-synNotch-TetR-VP64 (gift from Wendell Lim; Addgene plasmid # 79128; http://n2t.net/addgene:79128 ; RRID:Addgene_79128) (Morsut et al. 2016) along with an N- terminal myc epitope tag, and inserted into a linearized AAV-hSyn backbone. To reduce the ligand-independent activation observed in Ai63 TRE-TdTomato mice injected with AAV- Receiver only, we exchanged a SynNotchRAM7 containing a hydrophobic sequence present in the endogenous Notch (QHGQLWF) as done by Yang et al. (Yang et al. 2020). SynNotchRAM7 was synthesized commercially (Biobasic, Canada) and inserted into the pAAV.hSyn-LaG17-tTA vector. To create the pAAV.Grm6S-LaG17-SynNotch RAM7-tTA for selective expression in rod bipolar and ON cone bipolar cells, we inserted the promoter sequence containing four tandem repeats of a 200 bp element of the metabotropic glutamate receptor 6 gene (pAAV-Grm6S[4x]- tdT plasmid was a gift from Dr. Daniel Kerschensteiner) (Lagali et al. 2008; Tien et al. 2017). AAV administration Recombinant adeno-associated virus (rAAVs) for the TRACR system used in this study were produced as AAV2/9 or AAV2.7M8 viral particles (~1-5 x 1012-13 viral genome (vg)/mL) by Vigene BioSciences/Charles River (Rockville, MD, USA) or the Canadian Optogenetics and Vectorology Foundry (COVF) Viral Vector Core (Laval, Canada; RRID:SCR_016477). Viruses used and experiments are listed in Table S1. The following plasmids were obtained from Addgene (Watertown, MA, USA), and AAV2/9 vectors were produced by the COVF Viral Vector Core: pAAV-hSyn-DIO-tdTomato-2A-9xHA- Synaptophysin (a gift from Bernardo Sabatini, Addgene plasmid # 163686 ; http://n2t.net/addgene:163686 ; RRID:Addgene_163686) (Chantranupong et al. 2020); pAAV- TRE-mRuby2 (a gift from Viviana Gradinaru, Addgene plasmid # 99114 ; http://n2t.net/addgene:99114 ; RRID:Addgene_99114) (Chan et al. 2017). Intraocular injections of AAVs: AAVs were delivered into eyes of postnatal (P0-P4) or juvenile (P25 or older) mice by intraocular injection procedures described previously (Ing-Esteves et al. 2018; Gurdita et al. 2023). Postnatal pups were anesthetized by hypothermia by placing animals on ice until they were unresponsive to touch). Juvenile mice were anesthetised with 4% isoflurane and maintained on 2-2.5% isoflurane or with ketamine at 100 mg/kg of body weight for P20 and older. Animals were placed on a warming pad to maintain body temperature and fully recover from anesthesia before being returned to their cages. .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 8, 2026. ; https://doi.org/10.64898/2026.02.08.704659doi: bioRxiv preprint 21 For Sender and Receiver infections in the inner retina, intravitreal AAV delivery of juvenile mice, a 30-1/2 gauge needle (Becton Dickinson Mississauga, Canada) used to make a small hole just above the ora serrata, and 0.7-1.0 μl of rAAV mixed with Fast Green dye for visualisation was injected with a Hamilton syringe (7632-01) and 33 gauge blunt-ended needle (7803-05, Hamilton) into the intravitreal space. For Sender- and Receiver-AAV infection of photoreceptor or Müller glial cells in the outer retina, a subretinal injection was performed. Briefly, a 30G needle (BD, Mississauga, Canada) was used to make a small incision in the sclera at the nasal limbus, and the refluxed vitreous fluid was removed using Surgical spears (30-049, DeRoyal, Tennessee, US). A Hamilton syringe (80016, Chromatographic Specialties Inc., Brockville, Canada) with a 33G blunt needle was used to deliver 0.4-1 µl of the Sender AAV (1–5 × 1012-13 GC/ml) into the subretinal space, taking care to avoid lens damage and to achieve a localized subretinal bleb. For Receiver-AAV delivery to bipolar cells, an intravitreal injection of 0.5-1 µl of Receiver AAV at a similar titer was performed through the same incision using a similar Hamilton syringe. The animals were then placed on a heating pad (Life Brand, Toronto, ON, Canada) to fully recover from anesthesia before being returned to their cages. For tracing of local retinal circuits in Chat-Cre or Gad2-Cre mice, Sender AAV2/9.hSyn-DIO- (Nrx3b-GFP-2A-ChR2(H134R)-YFP), Receiver AAV2/9.hSyn-LaG17-SynNotch-tTA and Reporter AAV2/9.hSyn-TRE-mRuby reporter viruses were mixed at a 3:1:1 ratio. For Sender injections alone, hSyn-DIO-(Nrx3b-GFP-2A-ChR2(H134R)) was injected undiluted, Receiver/reporter injections included a 3:1 mix of undiluted receiver and TRE-mRuby2 diluted to 1012 GC/mL. When one component of the viral cocktail was excluded for controls, the remaining volume was made up with sterile PBS to preserve concentrations. Injections via stereotaxic surgeries were performed using a Stoelting stereotaxic apparatus with a model 62 RN Hamilton syringe driven by a syringe pump (#SP3101, World Precision Instruments, Sarasota, FL, USA). Mice were induced with 4% isoflurane and maintained on 2- 2.5% isoflurane throughout the operation. For testing TRACR along RGC projections, 4-5 week- old mice received simultaneous intravitreal administration of Sender-GFP AAV in retina (method as above) and stereotaxic delivery of Receiver-SynNotch and reporter AAVs to thalamus, and harvested 4-6 weeks post-injection. A 2 mm x 2 mm craniotomy was performed, and virus was injected into thalamus at AP -2.3 ML ±2.4 to 2.6 DV -2.9 to 3.0. 100 nL of virus mixture was injected at 25 nL/min to decrease viral spread. Receiver/reporter injections included a 3:1 mix of undiluted receiver and TRE-mRuby2 diluted to 1012 GC/mL. This allows for a final delivered TRE-reporter dose of 2.5 x 107 viral genomes, consistent with doses previously shown to .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 8, 2026. ; https://doi.org/10.64898/2026.02.08.704659doi: bioRxiv preprint 22 provide an optimal signal/leakiness ratio for retrograde transsynaptic tracing (Lavin et al. 2020). A dilution series of TRE reporter virus in the presence of receiver but without sender virus confirmed the absence of sender-independent background reporter activation at this dose in the thalamus (Supplemental figure 1). Two-photon guided electrophysiological recordings Two photon guided electrophysiological recordings were performed as previously described (Rochon et al. 2021; Olguin et al. 2025) was used to target retinal ganglion cells in retinal explants. Briefly, mice were dark adapted for at least 2 hr, euthanized, and then retinae rapidly dissected under infrared illumination into oxygenated (95% O2; 5% CO2) Ringer’s solution. Next, retinae were mounted onto a filter paper (MilliporeSigma, HABG01300) with the RGC layer facing up, placed in a recording chamber, mounted on the stage of a custom-built two-photon microscope, and perfused with oxygenated Ames solution warmed to 32–34°C. Patch electrodes (4–5 MΩ) were filled with Ringer’s solution, and fluorescein 3000 MW dextran (Thermo Scientific, D7156) was added to make the electrode visible under two-photon illumination. Signals were acquired with a MultiClamp 700B amplifier (Molecular Devices) and digitized at 20 kHz using custom software written in LabView. For spikes, the MultiClamp was put into I = 0 mode and Bessel filter set at 1 kHz. Analysis of electrophysiological signals was performed in MATLAB (Simulink) as follows. Briefly, action potentials were detected in loose patch recordings using the peakfinder function and binned (50 ms) over the entire length of the trial; firing rate histograms for each trial were then averaged and subjected to further processing based on each stimulus. Direction selective indices for ON, OFF, and ON-OFF RGCs were calculated using the circular variance of the cell response for all eight moving bar directions using standard methods described previously. Visual Stimuli A DLP light crafter (Texas Instruments, Dallas, TX) was used to project monochrome (410 nm) visual stimuli through a custom lens assembly that steered stimulus patterns into the back of a 20× objective (Euler et al. 2009). All visual stimuli were written in MATLAB using the psychophysics toolbox and displayed with a background intensity set to 1 × 104 R*/rod/s. Moving bar stimuli consisted of a bright bar moving along its long axis in one of eight directions. The bar was 200 μm wide, 1500 μm long moving at 1000μm/s. For electrophysiological experiments, the cell-receptive field center was identified using a grid of flashing spots and a .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 8, 2026. ; https://doi.org/10.64898/2026.02.08.704659doi: bioRxiv preprint 23 small user-controlled probe and the location with the highest response assigned as the center for all subsequent stimuli. Cortical neuron cultures: Cultures of dissociated cortices were prepared from timed pregnant females, similar to as previously described (Sengar et al. 2019). Briefly, E15 fetuses were decapitated and transferred to chilled Hank’s solution. The brain was removed and the meninges were peeled from each hemisphere. The cortices, without hippocampi, were pooled for each embryo and were mechanically dissociated using a P1000 pipette tip. The cells were plated on poly-D-lysine - coated glass coverslips in 12-well plates. The plating media was Neurobasal media (Gibco 21103-049) supplemented with fetal bovine serum (Gibco 26140), L-Glutamine (Gibco 25030- 081) and B27 (Gibco 17054-044). Cultures were maintained by replacing half the media with fresh maintenance media every 3-4 days, which was Neurobasal media supplemented with L- Glutamine, B27, and 1% Penicillin-Streptomycin (Sigma-Aldrich, P4333). For AAV transfections, DIV7 neurons were incubated with AAVs diluted to 1-2 x 109 GC/mL in maintenance medium for 7 days. At DIV14, neurons were fixed with 4% paraformaldehyde (PFA) + 4% sucrose for 15 minutes at room temperature. Tissue collection, immunohistochemistry Brain and retina tissues were harvested 4-6 weeks after AAV injections for tracing long-range projections, and 2-4 weeks for local retinal circuits, allowing time for TRACR expression, unless otherwise noted in Results. For brain tissue collection, mice were deeply anesthetised with isoflurane and transcardially perfused with normal saline (0.9% sodium chloride) or phosphate- buffered saline (PBS) followed by 4% paraformaldehyde in phosphate-buffered saline (PBS). Brains were dissected and post-fixed in 4% paraformaldehyde at 4 degrees overnight. Brains were then embedded in 4% agarose and sectioned at 75-150 µm with a VT1000S vibratome (Leica). Free-floating sections were stored at 4°C in PBS with 0.02% sodium azide until use. Whole-mount preparations and cryosections of retinas were prepared as described previously (Ing-Esteves et al. 2018). Eyes were removed from mice either euthanized then killed by decapitation, or transcardially perfused with PBS, and briefly fixed in ice-cold 4% paraformaldehyde. Retinas were dissected and post-fixed at 4°C for 2 hours or overnight. For some experiments, eyes were marked with a silver nitrate stick (118-395, AMG Medical Inc., Mont-Royal, Canada) on the dorsal side and then were carefully dissected from the eyecup. .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 8, 2026. ; https://doi.org/10.64898/2026.02.08.704659doi: bioRxiv preprint 24 Retinas for wholemounts were stored in PBS with 0.02% sodium azide until use. Retinas for cryosectioning were transferred to 30% sucrose in PBS overnight, embedded in Tissue-Tek OCT (Sakura Finetek), frozen in liquid nitrogen or on chilled methylbutane (Sigma). Tissues were cryosectioned at 16-20 µm onto Superfrost Plus glass slides (Thermo Fisher Scientific, Mississauga, Canada) using a cryostat (CM3050 S, Leica Biosystems, Richmond Hill, Canada) and dried for 2 hours before storage at -20 °C with desiccant in a slide box. For immunostainings of retina sections, slides were washed three times in PBS, then blocked and permeabilized with 2-5% donkey serum and 0.1-0.2% Triton X-100 in PBS (PBST). Floating brain sections were blocked with 5% donkey serum and 0.3% PBST. For stainings including primary antibodies raised in mouse, additional blocking of endogenous mouse IgG with was performed with AffiniPure Fab Fragment Donkey Anti-Mouse IgG (H+L) (#715-007-003, Jackson Immunoresearch) for 30 minutes at room temperature. Cryosections were probed with primary antibody in blocking buffer overnight, while floating sections were probed for three days at 4°C. After washing in PBST, samples were incubated with appropriate fluorophore- conjugated secondary antibodies in 2-5% NDS for 2-3 hours at room temperature. Where indicated, DAPI (1:5,000 from a 5 mg/mL solution, # D1306, Thermo Fisher Scientific) was included with the secondary antibodies. For cultured neurons, fixed cells were permeabilized with 0.25% Triton X-100 for 12 minutes and blocked in 5% NDS + 1% BSA in PBS for 1 hour. Cells were incubated with primary antibodies in blocking solution overnight at 4°C. Cell surface labelling was performed as previously described (Zhou et al. 2021). Briefly, fixed cells were blocked in 5% NDS + 1%BSA in PBS without permeabilization for 1 hour and then incubated with primary antibodies in blocking solution overnight at 4°C. Then, cells were permeabilized in 0.25% Triton X-100 for 10- 15 minutes, blocked again for 1 hour in blocking solution, and incubated with primary antibodies against intracellular antigens overnight at 4°C. Cells were then incubated with secondary antibodies in blocking solution for 2 hours at room temperature, followed by DAPI (1:10,000, ThermoFisher D1306) for 10 minutes. Coverslips were mounted on microscope slides (Fisherbrand, 22-034486) with Fluoromount-G mounting medium (Southern Biotech, 0100-01). Primary antibodies used were: Chicken anti-GFP (1:1000, Aves Labs, GFP-1010; RRID:AB_2307313); goat anti-GFP (1:500, Rockland, 600-101-215, RRID:AB_218182); mouse anti-myc (1:500, Millipore, 05-419; RRID:AB_309725); rabbit anti-myc (1:100-1:400, Cell Signaling Technology, 2272, RRID:AB_10692100); rabbit anti-RFP/DsRed (1:1000, Rockland, .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 8, 2026. ; https://doi.org/10.64898/2026.02.08.704659doi: bioRxiv preprint 25 600-401-379; RRID:AB_2209751); chicken anti-RFP (1:500, Rockland, 600-901-379, RRID: AB_10704808); goat anti-RFP (1:1000, Rockland, 200-101-379; RRID:AB_2744552); goat anti- ChAT (1:250, Millipore, AB144P; RRID:AB_2079751); rabbit anti-SATB1 (1:100, Abcam, ab49061; RRID:AB_882454); Hoechst (1:10000, Invitrogen, H3570); mouse anti-PKC (1:500, Santa Cruz Biotechnology, Sc-8393, RRID:AB_628142); rabbit anti-Ctbp2/Ribeye (1:10000, Synaptic Systems, 193-003, RRID:AB_2086768), mouse anti-PSD95 (1:500, Abcam, AB13552, RRID:AB_300453), rabbit anti-Opn4 (1:1000, Thermo Fisher Scientific, PA1-780, RRID:AB_2267547), mouse anti-SMI132 (1:1000, Covance, SMI-32P, RRID:AB_2314912), rabbit anti-Synapsin I (1:1000, Millipore, AB1543P, RRID:AB_90757), mouse anti-G0α (1:500, Millipore, MAB3073, RRID:AB_94671), rabbit anti SOX9 (1:500, Thermo Fisher Scientific, 702016, RRID:AB_2716879). Secondary antibodies conjugated to Dylight 405, Alexa Fluor 488, Alexa Fluor 568, Alexa Fluor 594 or Alexa Fluor 647 (Thermo Fisher Scientific or Jackson Immunoresearch) were used at 1:500 or 1:1000. Secondary antibodies used include: Donkey anti-Goat IgG 488 (1:500, Thermo Fisher Scientific, A11055, RRID:AB_2534102); Donkey anti-Mouse IgG 488 (1:500, Thermo Fisher Scientific, RRID:AB_141607); Donkey anti-Goat IgG 488 (1:1000, Jackson ImmunoResearch Labs, 705-545-147, RRID:AB_2336933); Donkey anti-Rabbit IgG 488 (1:1000, Jackson ImmunoResearch Labs, 711-545-152, RRID:AB_2313584); Donkey anti-Goat IgG 568 (1:500, Thermo Fisher Scientific, A11057, RRID:AB_2534104); Donkey anti-Chicken IgG 555(1:500, Thermo Fisher Scientific, A78950, RRID: AB_2921072); Donkey anti-Rabbit IgG 555 (1:500-1:1000, Thermo Fisher Scientific, A31572, RRID:AB_162543); Donkey anti-Rabbit IgG 568 (1:1000, Thermo Fisher Scientific, A10042, RRID:AB_2534017); Donkey anti-Guinea Pig IgG 633 (1:500, Sigma-Aldrich, SAB4600129, RRID:AB_2890636); Donkey anti-Rabbit IgG 647 (1:500, Thermo Fisher Scientific, A-31573, RRID: AB_2536183); Donkey anti-Rabbit IgG 647 (1:1000, Jackson ImmunoResearch Labs, 711-605-152, RRID:AB_2492288); Donkey anti- Goat IgG 647 (1:1000, Thermo Fisher Scientific, A-21447, RRID:AB_141844); Donkey anti- Mouse IgG 647 (1:1000, Jackson ImmunoResearch Labs, 715-605-151, RRID:AB_2340863). Image Acquisition and Analyses Immunofluorescence images were acquired using a Leica SP8, Leica Stellaris 5 (Leica Microsystems, Germany), LSM 780 or LSM810 (Carl Zeiss Inc., Thornwood, NY, USA) laser scanning confocal microscopes. Images were acquired with optimal xy- and z- resolutions defined by the objective’s numerical aperture and system settings, using 20X (NA=0.75), 40X oil (NA=1.3), or 63X glycerol (NA=1.3) lens. All the comparative images were taken using the same .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 8, 2026. ; https://doi.org/10.64898/2026.02.08.704659doi: bioRxiv preprint 26 microscope and acquisition parameters. ImageJ/Fiji (NIH) software was used for image processing and representations, including maximal projections and brightness/contrast enhancement (http://imagej.net/Fiji/Downloads; RRID:SCR_002285) (Schindelin et al. 2012). Laminar distributions of neurites within inner plexiform layer (IPL) were analysed using Fiji as previously described (Liu et al. 2018). Fluorescence intensity values were obtained from a line scan across the full IPL using the “Plot Profile” function. Localisation values were normalised such that the scleral border of the IPL represents 0% and the vitreal border of the IPL is indicated by 100%. Fluorescence intensity values were normalised to the minimum and maximum values within each line scan. Intensity values, in arbitrary units, were then arranged in 20 equal bins across the IPL. For retina TRACR-positive cell counts, the images were analyzed using Imaris 10 (Bitplane, Zurich, Switzerland). The spot tools in Imaris were used to count myc- Receiver, tdTomato-Reporter and marker-positive cells manually. The percentage of Reporter- positive Receiver cells or marker-positive reporter cells was determined by counting cells from one field per section, across three sections per retina, and in at least three retinas per condition. Quantifications and statistical analysis Number of experiments, animals and samples are reported in figure legends for each experiment and quantifications of TRACR cell labeling. Animals of either sex were analyzed. Statistical analyses were performed using Graph-Pad Prism software or the Real Statistics Resource Pack (Charles Zaiontz. www.real-statistics.com). All data are presented as mean ± s.e.m., unless otherwise stated. Means of two groups were compared using the two-tailed Student’s t test on condition of equivalent variances determined by the ANOVA F test, or with the Mann–Whitney nonparametric test. Means of multiple samples were compared using one- way ANOVA and Tukey’s multiple-comparisons test for pairwise analyses. To determine if ChAT-Cre and Gad2-Cre TRACR detected similar populations, similarity indices were calculated as previously described (Duan et al. 2018). Similarity indices were pooled by genotype and subjected to a one-way ANOVA to determine whether groups were significantly different. If differences were detected, posthoc pairwise tests were performed to determine the significance level reported in the figure legends. Exact p-values are reported unless the values are < 0.0001. .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 8, 2026. ; https://doi.org/10.64898/2026.02.08.704659doi: bioRxiv preprint 27

We thank the following investigators for generously sharing reagents: Dr. Liang Cai for plasmid sequence encoding SynNotch-RAM7; Dr. Daniel Kerschensteiner for the pAAV-Grm6S[4x]- tdTomato plasmid; Dr. Michel Cayouette for the R26LSL-TeNT mice, and Dr. Hongkui Zeng for the Ai63(TIT-tdTomato) mice. This work was supported by: Medicine by Design Canada First Research Excellence Fund (MbDNI-2020-01) to V.A.W. and J.L.L.; Fighting Blindness Canada Research Grant to J.L.L. and A.K.; NSERC Discovery Grant (RGPIN-2023-05107) to J.L.L.; CIHR Catalyst Grant (DV2- 197706 ) to J.L.L. and A.K. ; CIHR Project Grant (PJT195688) to V.A.W. and J.L.L; NIH award EY018139 to K.A.M. J.L.L. and A.K. were supported by a Tier 2 Canada Research Chair. V.A.W is supported by Donald K. Johnson Chair in Vision Research and Tier 1 Canada Research Chair. M.T.G. was supported by a CIHR Fellowship. M.I., S.M.E. and P.P. were supported by the University of Toronto Vision Science Research Scholarship. .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 8, 2026. ; https://doi.org/10.64898/2026.02.08.704659doi: bioRxiv preprint 28

Baden T et al. 2016. The func9onal diversity of re9nal ganglion cells in the mouse. Nature. 529(7586):345–350. hGps://doi.org/10.1038/nature16468 Bae JA et al. 2018. Digital Museum of Re9nal Ganglion Cells with Dense Anatomy and Physiology. Cell. 173(5):1293-1306.e19. hGps://doi.org/10.1016/j.cell.2018.04.040 Beier KT et al. 2013. Transsynap9c tracing with vesicular stoma99s virus reveals novel re9nal circuitry. J Neurosci. 33(1):35–51. hGps://doi.org/10.1523/JNEUROSCI.0245-12.2013 Beier KT. 2019. Hitchhiking on the neuronal highway: Mechanisms of transsynap9c specificity. J Chem Neuroanat. 99:9–17. hGps://doi.org/10.1016/j.jchemneu.2019.05.001 Bemben MA, Shipman SL, Nicoll RA, Roche KW. 2015. The cellular and molecular landscape of neuroligins. Trends Neurosci. 38(8):496–505. hGps://doi.org/10.1016/j.9ns.2015.06.004 Ben-Simon Y et al. 2025. A suite of enhancer AAVs and transgenic mouse lines for gene9c access to cor9cal cell types. Cell. 188(11):3045-3064.e23. hGps://doi.org/10.1016/j.cell.2025.05.002 Bouin A et al. 2024. New rabies viral resources for mul9-scale neural circuit mapping. Mol Psychiatry. 29(7):1951–1967. hGps://doi.org/10.1038/s41380-024-02451-6 Bowes C et al. 1990. Re9nal degenera9on in the rd mouse is caused by a defect in the beta subunit of rod cGMP-phosphodiesterase. Nature. 347(6294):677–680. hGps://doi.org/10.1038/347677a0 Briggman KL, Helmstaedter M, Denk W. 2011. Wiring specificity in the direc9on-selec9vity circuit of the re9na. Nature. 471(7337):183–188. hGps://doi.org/10.1038/nature09818 Brightman DS et al. 2016. Nrl-Cre transgenic mouse mediates loxP recombina9on in developing rod photoreceptors. Genesis. 54(3):129–135. hGps://doi.org/10.1002/dvg.22918 Brombas A, Kalita-de Croe S, Cooper-Williams EJ, Williams SR. 2017. Dendro-dendri9c cholinergic excita9on controls dendri9c spike ini9a9on in re9nal ganglion cells. Nat Commun. 8(1):15683. hGps://doi.org/10.1038/ncomms15683 Burris C et al. 2002. How Müller glial cells in macaque fovea coat and isolate the synap9c terminals of cone photoreceptors. J Comp Neurol. 453(1):100–111. hGps://doi.org/10.1002/cne.10397 Cachero S et al. 2020. BAcTrace, a tool for retrograde tracing of neuronal circuits in Drosophila. Nat Methods. 17(12):1254–1261. hGps://doi.org/10.1038/s41592-020-00989-1 Cao Y et al. 2015. Mechanism for Selec9ve Synap9c Wiring of Rod Photoreceptors into the Re9nal Circuitry and Its Role in Vision. Neuron. 87(6):1248–1260. hGps://doi.org/10.1016/j.neuron.2015.09.002 Chan KY et al. 2017. Engineered AAVs for efficient noninvasive gene delivery to the central and peripheral nervous systems. Nat Neurosci. 20(8):1172–1179. hGps://doi.org/10.1038/nn.4593 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 8, 2026. ; https://doi.org/10.64898/2026.02.08.704659doi: bioRxiv preprint 29 Chantranupong L et al. 2020. Rapid purifica9on and metabolomic profiling of synap9c vesicles from mammalian brain. Elife. 9:e59699. hGps://doi.org/10.7554/eLife.59699 Chen Q, Pei Z, Koren D, Wei W. 2016. S9mulus-dependent recruitment of lateral inhibi9on underlies re9nal direc9on selec9vity Feller MB, editor. eLife. 5:e21053. hGps://doi.org/10.7554/eLife.21053 Choe JH et al. 2021. SynNotch-CAR T cells overcome challenges of specificity, heterogeneity, and persistence in trea9ng glioblastoma. Sci Transl Med. 13(591):eabe7378. hGps://doi.org/10.1126/scitranslmed.abe7378 Coomer CE et al. 2025. Transsynap9c labeling and transcrip9onal control of zebrafish neural circuits. Nat Neurosci. 28(1):189–200. hGps://doi.org/10.1038/s41593-024-01815-z Daigle TL et al. 2018. A Suite of Transgenic Driver and Reporter Mouse Lines with Enhanced Brain-Cell- Type Targe9ng and Func9onality. Cell. 174(2):465-480.e22. hGps://doi.org/10.1016/j.cell.2018.06.035 Delerue F, White M, IGner LM. 2014. Inducible, 9ghtly regulated and non-leaky neuronal gene expression in mice. Transgenic Res. 23(2):225–233. hGps://doi.org/10.1007/s11248-013-9767-7 Ding H et al. 2016. Species-specific wiring for direc9on selec9vity in the mammalian re9na. Nature. 535(7610):105–110. hGps://doi.org/10.1038/nature18609 Duan X et al. 2018. Cadherin Combina9ons Recruit Dendrites of Dis9nct Re9nal Neurons to a Shared Interneuronal Scaffold. Neuron. 99(6):1145-1154.e6. hGps://doi.org/10.1016/j.neuron.2018.08.019 Euler T et al. 2009. Eyecup scope—op9cal recordings of light s9mulus-evoked fluorescence signals in the re9na. Pflugers Arch - Eur J Physiol. 457(6):1393–1414. hGps://doi.org/10.1007/s00424-008-0603-5 Fairless R et al. 2008. Polarized Targe9ng of Neurexins to Synapses Is Regulated by their C-Terminal Sequences. J Neurosci. 28(48):12969–12981. hGps://doi.org/10.1523/JNEUROSCI.5294-07.2008 Friedrichsen K et al. 2024. Subcellular pathways through VGluT3-expressing mouse amacrine cells provide locally tuned object-mo9on-selec9ve signals in the re9na. Nat Commun. 15(1):2965. hGps://doi.org/10.1038/s41467-024-46996-0 Furlanis E et al. 2025. An enhancer-AAV toolbox to target and manipulate dis9nct interneuron subtypes. bioRxiv. 2024.07.17.603924. hGps://doi.org/10.1101/2024.07.17.603924 Gokce O, Südhof TC. 2013. Membrane-Tethered Monomeric Neurexin LNS-Domain Triggers Synapse Forma9on. J Neurosci. 33(36):14617–14628. hGps://doi.org/10.1523/JNEUROSCI.1232-13.2013 Gradinaru V et al. 2010. Molecular and cellular approaches for diversifying and extending optogene9cs. Cell. 141(1):154–165. hGps://doi.org/10.1016/j.cell.2010.02.037 Graybuck LT et al. 2021. Enhancer viruses for combinatorial cell-subclass-specific labeling. Neuron. 109(9):1449-1464.e13. hGps://doi.org/10.1016/j.neuron.2021.03.011 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 8, 2026. ; https://doi.org/10.64898/2026.02.08.704659doi: bioRxiv preprint 30 Gurdita A et al. 2023. Progenitor division and cell autonomous neurosecre9on are required for rod photoreceptor sublaminar posi9oning. Proc Natl Acad Sci U S A. 120(42):e2308204120. hGps://doi.org/10.1073/pnas.2308204120 HaGar S et al. 2002. Melanopsin-containing re9nal ganglion cells: architecture, projec9ons, and intrinsic photosensi9vity. Science. 295(5557):1065–1070. hGps://doi.org/10.1126/science.1069609 Helmstaedter M et al. 2013. Connectomic reconstruc9on of the inner plexiform layer in the mouse re9na. Nature. 500(7461):168–174. hGps://doi.org/10.1038/nature12346 Hrva9n S et al. 2019. A scalable plaqorm for the development of cell-type-specific viral drivers. Elife. 8:e48089. hGps://doi.org/10.7554/eLife.48089 Huang T-H et al. 2017. Tracing neuronal circuits in transgenic animals by transneuronal control of transcrip9on (TRACT). Elife. 6:e32027. hGps://doi.org/10.7554/eLife.32027 Huang T-H, Velho T, Lois C. 2016. Monitoring cell-cell contacts in vivo in transgenic animals. Development. 143(21):4073–4084. hGps://doi.org/10.1242/dev.142406 Ing-Esteves S et al. 2018. Combinatorial Effects of Alpha- and Gamma-Protocadherins on Neuronal Survival and Dendri9c Self-Avoidance. J Neurosci. 38(11):2713–2729. hGps://doi.org/10.1523/JNEUROSCI.3035-17.2018 Ing-Esteves S, Lefebvre JL. 2024. Gamma-protocadherins regulate dendrite self-recogni9on and dynamics to drive self-avoidance. Curr Biol. 34(18):4224-4239.e4. hGps://doi.org/10.1016/j.cub.2024.08.002 Jiang L et al. 2004. Tight regula9on from a single tet-off rAAV vector as demonstrated by flow cytometry and quan9ta9ve, real-9me PCR. Gene Ther. 11(13):1057–1067. hGps://doi.org/10.1038/sj.gt.3302245 Jin L et al. 2024. Long-term labeling and imaging of synap9cally connected neuronal networks in vivo using double-dele9on-mutant rabies viruses. Nat Neurosci. 27(2):373–383. hGps://doi.org/10.1038/s41593-023-01545-8 JüGner J et al. 2019. Targe9ng neuronal and glial cell types with synthe9c promoter AAVs in mice, non- human primates and humans. Nat Neurosci. 22(8):1345–1356. hGps://doi.org/10.1038/s41593-019- 0431-2 Kim J et al. 2011. mGRASP enables mapping mammalian synap9c connec9vity with light microscopy. Nat Methods. 9(1):96–102. hGps://doi.org/10.1038/nmeth.1784 KlaG O et al. 2021. Endogenous β-neurexins on axons and within synapses show regulated dynamic behavior. Cell Rep. 35(11):109266. hGps://doi.org/10.1016/j.celrep.2021.109266 Krashes MJ et al. 2011. Rapid, reversible ac9va9on of AgRP neurons drives feeding behavior in mice. J Clin Invest. 121(4):1424–1428. hGps://doi.org/10.1172/JCI46229 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 8, 2026. ; https://doi.org/10.64898/2026.02.08.704659doi: bioRxiv preprint 31 Krishnaswamy A et al. 2015. SIDEKICK 2 DIRECTS FORMATION OF A RETINAL CIRCUIT THAT DETECTS DIFFERENTIAL MOTION. Nature. 524(7566):466–470. hGps://doi.org/10.1038/nature14682 Lagali PS et al. 2008. Light-ac9vated channels targeted to ON bipolar cells restore visual func9on in re9nal degenera9on. Nat Neurosci. 11(6):667–675. hGps://doi.org/10.1038/nn.2117 Lavin TK, Jin L, Lea NE, Wickersham IR. 2020. Monosynap9c Tracing Success Depends Cri9cally on Helper Virus Concentra9ons. Front Synap9c Neurosci. 12:6. hGps://doi.org/10.3389/fnsyn.2020.00006 Lefebvre JL et al. 2012. Protocadherins mediate dendri9c self-avoidance in the mammalian nervous system. Nature. 488(7412):517–521. hGps://doi.org/10.1038/nature11305 Li E et al. 2021. Anterograde transneuronal tracing and gene9c control with engineered yellow fever vaccine YFV-17D. Nat Methods. 18(12):1542–1551. hGps://doi.org/10.1038/s41592-021-01319-9 Liu J et al. 2018. Tbr1 instructs laminar paGerning of re9nal ganglion cell dendrites. Nat Neurosci. 21(5):659–670. hGps://doi.org/10.1038/s41593-018-0127-z Lo L, Anderson DJ. 2011. A Cre-dependent, anterograde transsynap9c viral tracer for mapping output pathways of gene9cally marked neurons. Neuron. 72(6):938–950. hGps://doi.org/10.1016/j.neuron.2011.12.002 Madisen L et al. 2015. Transgenic mice for intersec9onal targe9ng of neural sensors and effectors with high specificity and performance. Neuron. 85(5):942–958. hGps://doi.org/10.1016/j.neuron.2015.02.022 Malagu9 M et al. 2022. SyNPL: Synthe9c Notch pluripotent cell lines to monitor and manipulate cell interac9ons in vitro and in vivo. Development. 149(12):dev200226. hGps://doi.org/10.1242/dev.200226 Martell JD et al. 2016. A split horseradish peroxidase for detec9on of intercellular protein-protein interac9ons and sensi9ve visualiza9on of synapses. Nat Biotechnol. 34(7):774–780. hGps://doi.org/10.1038/nbt.3563 Martersteck EM et al. 2017. Diverse Central Projec9on PaGerns of Re9nal Ganglion Cells. Cell Rep. 18(8):2058–2072. hGps://doi.org/10.1016/j.celrep.2017.01.075 Matsuda T, Cepko CL. 2004. Electropora9on and RNA interference in the rodent re9na in vivo and in vitro. Proc Natl Acad Sci U S A. 101(1):16–22. hGps://doi.org/10.1073/pnas.2235688100 Mauss AS, Vlasits A, Borst A, Feller M. 2017. Visual Circuits for Direc9on Selec9vity. Annu Rev Neurosci. 40:211–230. hGps://doi.org/10.1146/annurev-neuro-072116-031335 Morsut L et al. 2016. Engineering Customized Cell Sensing and Response Behaviors Using Synthe9c Notch Receptors. Cell. 164(4):780–791. hGps://doi.org/10.1016/j.cell.2016.01.012 Olguin AGR et al. 2025. Cadherin 4 assembles a family of color-preferring re9nal circuits that respond to light offset. Curr Biol. 35(6):1298-1310.e7. hGps://doi.org/10.1016/j.cub.2025.02.008 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 8, 2026. ; https://doi.org/10.64898/2026.02.08.704659doi: bioRxiv preprint 32 Osakada F, Callaway EM. 2013. Design and genera9on of recombinant rabies virus vectors. Nat Protoc. 8(8):1583–1601. hGps://doi.org/10.1038/nprot.2013.094 Osterwalder M et al. 2010. Dual RMCE for efficient re-engineering of mouse mutant alleles. Nat Methods. 7(11):893–895. hGps://doi.org/10.1038/nmeth.1521 Peng Y-R et al. 2017. Satb1 Regulates Contac9n 5 to PaGern Dendrites of a Mammalian Re9nal Ganglion Cell. Neuron. 95(4):869-883.e6. hGps://doi.org/10.1016/j.neuron.2017.07.019 Petrin D et al. 2003. Structural and func9onal protec9on of photoreceptors from MNU-induced re9nal degenera9on by the X-linked inhibitor of apoptosis. Invest Ophthalmol Vis Sci. 44(6):2757–2763. hGps://doi.org/10.1167/iovs.02-0729 Prigge CL et al. 2023. Rejec9on of inappropriate synap9c partners in mouse re9na mediated by transcellular FLRT2-UNC5 signaling. Dev Cell. 58(20):2080-2096.e7. hGps://doi.org/10.1016/j.devcel.2023.07.011 Reddy NR et al. 2024. Engineering synthe9c suppressor T cells that execute locally targeted immunoprotec9ve programs. Science. 386(6726):eadl4793. hGps://doi.org/10.1126/science.adl4793 Rivera JF et al. 2025. ATLAS: a ra9onally designed anterograde transsynap9c tracer. Nat Methods. 22(5):1101–1111. hGps://doi.org/10.1038/s41592-025-02670-x Rochon P-L, Theriault C, Rangel Olguin AG, Krishnaswamy A. 2021. The cell adhesion molecule Sdk1 shapes assembly of a re9nal circuit that detects localized edges Feller MB, Westbrook GL, editors. eLife. 10:e70870. hGps://doi.org/10.7554/eLife.70870 Rossi J et al. 2011. Melanocor9n-4 receptors expressed by cholinergic neurons regulate energy balance and glucose homeostasis. Cell Metab. 13(2):195–204. hGps://doi.org/10.1016/j.cmet.2011.01.010 Roybal KT et al. 2016. Engineering T Cells with Customized Therapeu9c Response Programs Using Synthe9c Notch Receptors. Cell. 167(2):419-432.e16. hGps://doi.org/10.1016/j.cell.2016.09.011 Schindelin J et al. 2012. Fiji: an open-source plaqorm for biological-image analysis. Nat Methods. 9(7):676–682. hGps://doi.org/10.1038/nmeth.2019 Sengar AS et al. 2019. Control of Long-Term Synap9c Poten9a9on and Learning by Alterna9ve Splicing of the NMDA Receptor Subunit GluN1. Cell Reports. 29(13):4285-4294.e5. hGps://doi.org/10.1016/j.celrep.2019.11.087 Shekhar K et al. 2016. Comprehensive Classifica9on of Re9nal Bipolar Neurons by Single-Cell Transcriptomics. Cell. 166(5):1308-1323.e30. hGps://doi.org/10.1016/j.cell.2016.07.054 Talay M et al. 2017. Transsynap9c Mapping of Second-Order Taste Neurons in Flies by trans-Tango. Neuron. 96(4):783-795.e4. hGps://doi.org/10.1016/j.neuron.2017.10.011 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 8, 2026. ; https://doi.org/10.64898/2026.02.08.704659doi: bioRxiv preprint 33 Taniguchi H et al. 2011. A resource of Cre driver lines for gene9c targe9ng of GABAergic neurons in cerebral cortex. Neuron. 71(6):995–1013. hGps://doi.org/10.1016/j.neuron.2011.07.026 Tien N-W, Soto F, Kerschensteiner D. 2017. Homeosta9c Plas9city Shapes Cell-Type-Specific Wiring in the Re9na. Neuron. 94(3):656-665.e4. hGps://doi.org/10.1016/j.neuron.2017.04.016 Toda S et al. 2018. Programming self-organizing mul9cellular structures with synthe9c cell-cell signaling. Science. 361(6398):156–162. hGps://doi.org/10.1126/science.aat0271 Tran NM et al. 2019. Single-Cell Profiles of Re9nal Ganglion Cells Differing in Resilience to Injury Reveal Neuroprotec9ve Genes. Neuron. 104(6):1039-1055.e12. hGps://doi.org/10.1016/j.neuron.2019.11.006 Tsai NY et al. 2022. Trans-Seq maps a selec9ve mammalian re9notectal synapse instructed by Nephronec9n. Nat Neurosci. 25(5):659–674. hGps://doi.org/10.1038/s41593-022-01068-8 Tsetsenis T et al. 2014. Direct visualiza9on of trans-synap9c neurexin-neuroligin interac9ons during synapse forma9on. J Neurosci. 34(45):15083–15096. hGps://doi.org/10.1523/JNEUROSCI.0348-14.2014 Viney TJ et al. 2007. Local re9nal circuits of melanopsin-containing ganglion cells iden9fied by transsynap9c viral tracing. Curr Biol. 17(11):981–988. hGps://doi.org/10.1016/j.cub.2007.04.058 Viswanathan S et al. 2015. High-performance probes for light and electron microscopy. Nat Methods. 12(6):568–576. hGps://doi.org/10.1038/nmeth.3365 Vong L et al. 2011. Lep9n ac9on on GABAergic neurons prevents obesity and reduces inhibitory tone to POMC neurons. Neuron. 71(1):142–154. hGps://doi.org/10.1016/j.neuron.2011.05.028 Wang B, Zhang Y. 2023. Asymmetric connec9ons with starburst amacrine cells underlie the upward mo9on selec9vity of J-type re9nal ganglion cells. PLoS Biol. 21(9):e3002301. hGps://doi.org/10.1371/journal.pbio.3002301 Wickersham IR, Lyon DC, et al. 2007. Monosynap9c Restric9on of Transsynap9c Tracing from Single, Gene9cally Targeted Neurons. Neuron. 53(5):639–647. hGps://doi.org/10.1016/j.neuron.2007.01.033 Wickersham IR, Finke S, Conzelmann K-K, Callaway EM. 2007. Retrograde neuronal tracing with a dele9on-mutant rabies virus. Nat Methods. 4(1):47–49. hGps://doi.org/10.1038/nmeth999 Williams PR et al. 2010. In vivo development of outer re9nal synapses in the absence of glial contact. J Neurosci. 30(36):11951–11961. hGps://doi.org/10.1523/JNEUROSCI.3391-10.2010 Xiong F et al. 2022. An HSV-1-H129 amplicon tracer system for rapid and efficient monosynap9c anterograde neural circuit tracing. Nat Commun. 13(1):7645. hGps://doi.org/10.1038/s41467-022- 35355-6 Xu X et al. 2020. Viral Vectors for Neural Circuit Mapping and Recent Advances in Trans-synap9c Anterograde Tracers. Neuron. 107(6):1029–1047. hGps://doi.org/10.1016/j.neuron.2020.07.010 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 8, 2026. ; https://doi.org/10.64898/2026.02.08.704659doi: bioRxiv preprint 34 Yamagata M, Sanes JR. 2012. Transgenic strategy for iden9fying synap9c connec9ons in mice by fluorescence complementa9on (GRASP). Front Mol Neurosci. 5:18. hGps://doi.org/10.3389/fnmol.2012.00018 Yan W et al. 2020. Mouse Re9nal Cell Atlas: Molecular Iden9fica9on of over Sixty Amacrine Cell Types. J Neurosci. 40(27):5177–5195. hGps://doi.org/10.1523/JNEUROSCI.0471-20.2020 Yang Z-J et al. 2020. Engineering of an enhanced synthe9c Notch receptor by reducing ligand- independent ac9va9on. Commun Biol. 3(1):116. hGps://doi.org/10.1038/s42003-020-0848-x Yonehara K et al. 2011. Spa9ally asymmetric reorganiza9on of inhibi9on establishes a mo9on-sensi9ve circuit. Nature. 469(7330):407–410. hGps://doi.org/10.1038/nature09711 Yoshida K et al. 2001. A Key Role of Starburst Amacrine Cells in Origina9ng Re9nal Direc9onal Selec9vity and Optokine9c Eye Movement. Neuron. 30(3):771–780. hGps://doi.org/10.1016/S0896- 6273(01)00316-6 Zhang S et al. 2022. Monitoring of cell-cell communica9on and contact history in mammals. Science. 378(6623):eabo5503. hGps://doi.org/10.1126/science.abo5503 Zhang Y et al. 2008. V3 spinal neurons establish a robust and balanced locomotor rhythm during walking. Neuron. 60(1):84–96. hGps://doi.org/10.1016/j.neuron.2008.09.027 Zhao S et al. 2008. Improved expression of halorhodopsin for light-induced silencing of neuronal ac9vity. Brain Cell Biol. 36(1–4):141–154. hGps://doi.org/10.1007/s11068-008-9034-7 Zhou Z et al. 2021. NGPF2 triggers synap9c scaling up through ALK-LIMK-cofilin-mediated mechanisms. Cell Rep. 36(7):109515. hGps://doi.org/10.1016/j.celrep.2021.109515 Zingg B et al. 2017. AAV-Mediated Anterograde Transsynap9c Tagging: Mapping Cor9cocollicular Input- Defined Neural Pathways for Defense Behaviors. Neuron. 93(1):33–47. hGps://doi.org/10.1016/j.neuron.2016.11.045 Zingg B et al. 2020. Synap9c Specificity and Applica9on of Anterograde Transsynap9c AAV for Probing Neural Circuitry. J Neurosci. 40(16):3250–3267. hGps://doi.org/10.1523/JNEUROSCI.2158-19.2020 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 8, 2026. ; https://doi.org/10.64898/2026.02.08.704659doi: bioRxiv preprint RFP SENDER RECEIVER REPORTER 1. 2. 3. Pro WPRE pA hSyn myc-LaG17 synNotch tTAhSyn myc-LaG17 synNotch-RAM7hSyn ChR2-eYFPT2AGFP Nrxn3β TRE (tetO) RFP WPRE pA SENDER AAV RECEIVER AAV REPORTER AAV or transgenic mouse, Ai63-TIT-TdTomato tTA ChR2-eYFPT2A GFPNrxn3β WPRE pA hSyn Flpo T2A GFPNrxn3β WPREpA WPRE pA WPRE pA dLGN VPM dLGN LD post VPMvLGN AAV-DIO-HASV- 2A-TdTomato A B C E H D TRE AAV-SENDER-GFP AAV-RECEIVER GFP (-TX) GFP (-TX)/ synapsin1 MYC/DAPI (+TX) MYC/DAPI (-TX) tTA AAV-DIO- SENDER-GFP AAV-DIO SENDER-GFP AAV-RECEIVER AAV-REPORTER vGlut2-Cre vGlut2-CredLGN dLGN Figure 1 TdTomato; Sender GFPTdTomato HA-SV; Sender F Lorem ipsum GFPHAG Imerge merge Figure 1: TRACR system and transneuronal tracing of long-range connections. (A) Schematic of TRACR. A presynaptic Sender neuron (green) displays an NRX–GFP ligand enriched at synaptic terminals. A postsynaptic Receiver neuron (magenta) expresses synNotch receptors comprising an extracellular GFP nanobody (LaG17), the Notch core regulatory and transmembrane domains (grey), and the intracellular transcriptional activator tTA. Ligand binding across the synapse [1] triggers intramembrane proteolysis of synNotch [2] and release of tTA, which translocates to the nucleus to activate a TRE-responsive Reporter [3], labeling the postsynaptic neuron with RFP. (B) TRACR components. Sender AAVs encode NRX-GFP, a 2A peptide, and a marker or effector transgene. Sender expression is driven by either a cell-type-specific promoter (Pro) or a pan-neuronal promoter (hSyn), and can be restricted by Cre-dependent DIO. Receiver AAVs encode two synNotch-tTA variants with an N-terminal Myc. TRE-driven reporters used here include AAV-TRE-mRuby2 and Ai63 (TIT-tdTomato) mice. .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 8, 2026. ; https://doi.org/10.64898/2026.02.08.704659doi: bioRxiv preprint Figure 1 (cont’d): TRACR system and transneuronal tracing of long-range connections. (C) Sender-GFP localization in cultured DIV14 cortical neurons co-infected with AAV-hSyn-Cre and AAV-hSyn-DIO-NRXGFP-T2A-Flpo. Surface staining (green; no Triton X-100 (-TX) shows NRX–GFP at the plasma membrane and at presynaptic sites overlapping with Synapsin1 (red, with sequential permeabilization; inset enlarged at right). (D) synNotch localization in cultured neurons infected with Receiver AAV-hSyn-myc-LaG17-synNotch-tTA. Immunostaining for Myc (white) shows synNotch along neurites and at the cell surface (right, no Triton X-100). (E) Strategy to visualize Sender-GFP within presynaptic terminals of retinal ganglion cells (RGCs) in dorsal lateral geniculate nuclei (dLGN). AAVs encoding Sender (hSyn-DIO-NRXGFP-T2A-Flpo) and synaptic vesicle marker (hSyn-DIO-HA-SV-2A-tdTomato) were delivered to retinas of vGluT2-Cre mice. (F) Confocal image of dLGN (dashed outline) shows NRX-GFP (green) overlap with RGC terminal axons (TdTomato, magenta). Inverted images of tdTomato (middle) and GFP channels (right) show lack GFP along TdTomato+ axon tracts (black arrows). (G) Localization of NRX-GFP ligand (green) within RGC terminal boutons, and overlapping with HA-tagged synaptic vesicle (SV) marker (magenta). (H) AAV strategy to express Sender (hSyn-DIO-NRXGFP-T2A-ChR2-YFP) in the retina, and Receiver (hSyn-synNotch) and Reporter (TRE-mRuby2) in the thalamus of vGluT2-Cre mice. Mice were injected at 3-4 postnatal weeks and harvested at 2 months of age. (I) Section of thalamus showing TRACR labeled retinothalamic connections. Receiver neurons (Myc, orange) are broadly distributed across thalamic nuclei, whereas Reporter expression (RFP, magenta) is restricted to Myc+ Receiver cells within the target region (dLGN) and is absent from non-target nuclei, including laterodorsal (LD) and ventral posteromedial (VPM). Right: enlarged insets. Scale bars: 50 µm (C,D); 200 µm (F, top) and 20µm (F, bottom); 200 µm (H, left) and 10 µm (H, right). .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 8, 2026. ; https://doi.org/10.64898/2026.02.08.704659doi: bioRxiv preprint AC SAC IPL thickness DAPI Sender-GFP Reporter Sender AAV: hSyn-DIO- NrxGFP-T2A-ChR2-YFP Chat-Cre x DIO-Sender Gad2-Cre x DIO-Sender Sender+ SACs Receiver Reporter+ SACs DSGCs Sender+ GABAergic ACs Receiver Reporter+ ACs RGCs G C PhR RGC MG BC Gad2-Cre Sender Reporter ChAT+SATB1 Reporter OPN4Reporter ChAT-Cre Sender AAV-Sender-GFP AAV-TRE-RFP ONL INL IPL GCL A E 25 50 75 100 0.0 0.5 1.0 1.5 IPL thickness S2 S4 Intensity (a.u.) 25 50 75 100 0 0.5 1.0 Intensity (a.u.) Intensity (a.u.) Sender-GFP Reporter D Pax6+SATB1+SATB1+ ChAT+SMI-32+OPN4+ 100% 50% 0% Percent of Marker+ Reporter+ cells Chat-Cre Gad2-Cre S2 S4 B F H I J K Figure 2 AAV-Receiver (myc) AAV-TRE-RFP HZ ONL OPL INL IPL GCL Receiver AAV: hSyn- MycLaG17-synNotch-tTA Reporter: AAV-TRE-RFP .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 8, 2026. ; https://doi.org/10.64898/2026.02.08.704659doi: bioRxiv preprint Figure 2: TRACR identifies interneuron targets in the inner retina. (A) Schematic of the retina. Photoreceptors (PhR) in the outer nuclear layer (ONL) synapse in the outer plexiform layer (OPL) onto bipolar cell (BP) and horizontal cell (HZ) dendrites. BPs and amacrine cells (ACs) project to the inner plexiform layer (IPL), where they form stereotyped, lamina-specific connections with retinal ganglion cell (RGC) subtypes. Starburst amacrine cells (SACs, green) are inhibitory ACs that stratify in two IPL sublaminae and synapse onto other SACs and direction-selective RGCs (DSGCs). (B) AAV TRACR components and controls. Sender, Receiver, and TRE-driven Reporter. Mice received AAV intravitreal injections at 3-4 postnatal weeks and harvested at 2 months of age. ChAT-Cre retina injected with Receiver and Reporter AAVs but lacking Sender shows Receiver expression (Myc, white) without TRE-driven RFP (middle). Retinas injected with Sender and Reporter AAVs shows GFP ligand (green) but no RFP induction (right). (C) Schematic of TRACR strategy to label SAC targets. In ChAT-Cre retinas, Sender (green) is rstricted to SACs (green), whereas Rec and Rep are broadly expressed across inner retinal neurons (grey). Reporter induction (magenta) is predicted in SACs and DSGCs. (D) Representative section of TRACR-labeled ChAT-Cre retina. Sender-GFP (green) and Reporter+ neurites (RFP, magenta) co-stratify in the expected IPL sublaminae. Right: Insets highlight overlapping neurites (yellow arrowheads). (E) Quantification of Sender-positive and Reporter-positive neurite stratification in ChAT-Cre retinas. Line profiles of Sender-GFP and RFP fluorescence across the IPL show strong anatomical correspondence (cos θ similarity index = 0.89 ± 0.05). Data are shown as mean (dark line) ± SEM (light ribbon); n = 3 images per retina, 4 retinas from 4 animals. (F) TRACR strategy to label GABAergic AC targets. In Gad2-Cre retinas, Sender (GFP, green) is expressed across broad AC populations. Reporter induction (RFP, magenta) is predicted in diverse ACs and RGCs. (G) Representative section of TRACR-labeled Gad2-Cre retina. Sender+ (green) and Reporter+ neurites (magenta) are throughout the IPL. (H) Line profiles of Sender+ and RFP+ neurites in Gad2-Cre retinas overlap (cos θ similarity index = 0.93 ± 0.07) but show distinct distributions from ChAT-Cre retinas in F. Data are shown as mean (dark line) ± SEM (light ribbon); n = 3 images per retina, 4 retinas from 4 animals. Similarity indices differ significantly from F, p < 0.01, one-way ANOVA. (I) Molecular validation of TRACR targets in ChAT-Cre retinas. Reporter-labeled cells (magenta) include SACs (ChAT, yellow; arrowheads) and DSGCs (Satb1, yellow; arrows) and exclude non-target RGC types such as melanopsin-positive ipRGCs (Opn4, yellow in K). (J) Molecular validation of TRACR targets in Gad2-Cre retinas using same markers as in (J). (K) TRACR targets labeled by Chat-Cre-driven Senders differ from Gad2-Cre-driven Senders. Plots show the percentage of marker-positive Reporter-positive neurons for Pax6 (pan ACs and RGCs), SATB1 (ON–OFF DSGCs), SATB1+ChAT (DSGCs/SACs), SMI-32 (α-RGCs), and Opn4 (ipRGCs). 3–5 retinas per condition. p = 0.64 (Pax6), 0.034 (SATB1), 0.0015 (SATB1+ChAT), 0.046 (SMI-32), 0.037 (Opn4). Scale bars: 50 µm (B, D, G, J, K). .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 8, 2026. ; https://doi.org/10.64898/2026.02.08.704659doi: bioRxiv preprint ON RFP+ ON-OFF RGCs OFF 0 0.1 0.2 0.3 0.4DSI ON OFF 1s F 0 0.2 0.4 0.6 0.8Fractino of RGCs ON OFF ON OFF RFP+ RFP- C D G 0° 90° 80Hz 0° 90Hz D N V T RFP+ RFP- RFPA488 D N V T E B Lectin/MycRFP 1mm (-) (-) (-) null preferred A Figure 3. Functional recordings of TRACR-positive cells confirm direction-selective identity. (A) Schematic of TRACR strategy showing Sender+ starburst amacrine cells (green), Receiver+ ACs and RGCs (grey) and Reporter labelled DSGCs (RFP, magenta). Recorded RFP cells are predicted to show direction-selective responses to preferred direction (black). Mice received AAV intravitreal injections at 3-4 postnatal weeks and harvested at 2 months of age; AAV-TRE-mRuby and Ai63 mice were used for Reporters. (B) Wholemount retina stained with antibodies against RFP, isolectin to label blood vessels, and anti-Myc to mark Receiver-labelled retinal neurons. High power image at right shows a typical field. (C) Fraction of RFP-positive (+) and RFP-negative (-) RGCs with ON, OFF, and ON-OFF responses to full field flash stimuli. (D) Two-photon image of alexa488 filled electrode (A488) and RFP+ RGC (top) and its spike responses evoked by a full field flash (bottom). (E) Polar plot of firing evoked by a bright bar moving in 8 directions. ON responses evoked by the leading edge of the bar, OFF responses evoked by the trailing edge. Normalized firing vector showing direction selective index (DSI) and angular preference also shown. (F) Average ON and OFF DSIs computed from recordings from RFP+ (n=32 RGCs) and RFP- cells (n=42 RGCs), like shown in D. Data show mean ± SEM. p < 0.05, t-test. (G) Polar plot showing angular preference from RFP+ ON-OFF RGCs. Data come from recordings in over 30 retinas from 25 mice. Scale bars: 1mm (B), 20 µm (D). Figure 3 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 8, 2026. ; https://doi.org/10.64898/2026.02.08.704659doi: bioRxiv preprint Sender GFP: AAV-ProC1-NrxGFP Receiver-myc: AAV-4xGrm6-MycLaG17- SynNotchRAM7-tTA Hoechst Sender PSD95 RIBEYE A B Hoechst Sender Receiver ReporterReceiver only C Sender + Receiver %Reporter+ Rec cells R S+Rec Hoechst Sender Receiver Reporter Hoechst PKC SCGN ReporterD PhR Sender + PhR Rec MG Sender + BP Rec Hoechst Sender Receiver Reporter Hoechst Sender Receiver ReporterE ONL OPL INL ONL OPL INL IPL ONL OPL INL IPL ONL OPL INL IPL ONL OPL INL IPL ONL OPL INL IPL % Marker+ Reporter+ cells Ai63 Reporter mice ONL OPL INL IPL ONL OPL INL IPL ONL OPL INL IPL %Reporter+ Rec Cells PhR MG P5 P20 CBP RBPCBP;RBP Figure 4 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 8, 2026. ; https://doi.org/10.64898/2026.02.08.704659doi: bioRxiv preprint Figure 4. TRACR reports synaptically connected cells in the photoreceptor–bipolar cell circuit. (A) Schematic of the TRACR strategy and predicted activation pattern at the photoreceptor–bipolar cell synapse. At postnatal day (P) 5, mice were co-infected with a Sender AAV driving NRX–GFP expression in photoreceptors (subretinal delivery of AAV-ProC1-NRXGFP-T2A-JawsHA), and an Receiver-AAV (intravitreal delivery of AAV-4xGrm6-MycLaG17-synNotchRAM7-tTA) driving synNotch expression in bipolar cells of Ai63 reporter mice. Synaptic contact between Sender- and Receiver-expressing cells induces TdTomato (RFP) reporter expression in Receiver-positive cells, analyzed at P20. (B) Localization of the NRX–GFP ligand in photoreceptor terminals. Retinal cross-section showing NRX–GFP fluorescence (without GFP amplification) enrichment in photoreceptor terminals, in proximity to presynaptic markers Ribeye (magenta) and PSD95 (white). (C) TRACR activation requires Sender and Receiver expression. Representative P20 retinal sections from Ai63 mice infected with the indicated AAVs and stained for Sender (GFP, green), Receiver (Myc, white), Reporter (TdTomato, magenta), and nuclei (Hoechst, blue). Left: Receiver (Rec)-only controls show minimal ligand-independent reporter activation). Middle: Co-injection of Sender (S) and Receiver results in robust reporter activation confined to Receiver-expressing bipolar cells (middle). Right: Percentage of Reporter-positive Receiver-expressing cells. Each data point represents the mean of three sections from a single retina, 6 retinas per condition. Data are shown as mean ± SD. ****p < 0.0001, Welch's t-test. (D) TRACR detects connections to rod and cone bipolar cells. Left: Sender- and Receiver-infected retinas from P20 Ai63 mice show rod bipolar cells (RBC; PKC, yellow), and cone bipolar cells (CBC, SCGN, cyan) express RFP (magenta). Right: Percentage of marker-positive Reporter-positive cells. N= 6 retinas. Data are shown as mean ± SD. (E) Ligand proximity alone is insufficient to activate TRACR. Left: Schematic and representative retinal sections showing photoreceptor (PhR)-driven Sender (GFP, green; AAV-ProC1-NRXGFP) and Receiver expression (Myc, white; AAV-ProC1-synNotch). Middle: Schematic and representative sections showing Müller glia-driven Sender (AAV-ProB2-NRXGFP) and Receiver expression (AAV-Grm6-synNotch) in bipolar cells. In both conditions, reporter activation is minimal or absent. Right: Percentage of Reporter-positive Receiver-expressing cells. N = 3 retinas (PhR, Photoreceptor Sender/Receiver), 6 retinas (MG, Müller glia Sender). Data are shown as mean ± SD. Scale bars: 5 μm (B), 20 μm (C, D, E). .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 8, 2026. ; https://doi.org/10.64898/2026.02.08.704659doi: bioRxiv preprint IPL IPL Hoechst Sender Receiver Reporter Control B ONL OPL INL IPL ONL OPL INL Hoechst Sender Receiver Reporter OPL INL IPL MNU C A Hoechst Sender Receiver Reporter Hoechst Sender Receiver Reporter rd1 P12 rd1 P30 INL OPL Early rd1; Ai63 or Control P12 %Reporter+ Rec cells P30 P60 rd1 %Reporter+ Rec cells Cont MNU PhR Sender ON BP Receiver rd1; Ai63 or MNU Figure 5 Figure 5. TRACR activation requires photoreceptor survival. (A) Schematic of the TRACR strategy in rd1;Ai63 and MNU-treated Ai63 mice. Mice were infected at P5 with AAV-PhR-Sender and AAV-BP-Receiver (as in Figure 4A) and analyzed for reporter induction prior to and after photoreceptor degeneration. (B) TRACR activation is progressively lost during photoreceptor degeneration in rd1 retina. Representative retinal sections from TRACR infected rd1;Ai63 mice and stained for Sender (GFP, green), Receiver (Myc, white), Reporter (TdTomato, magenta), and nuclei (Hoechst, blue) at P12 and P30, corresponding to early and late-stage rod degeneration. Right: Percentage of Reporter-positive Receiver-expressing cells at the indicated time points. Each data point represents the mean of three sections from a single retina. N = 3 retinas (P12), 7 retinas (P30), 5 retinas (P60). Data are shown as mean ± SD. p < 0.0001, one-way ANOVA, with Tukey pairwise comparison. (C) TRACR activation is lost in an acute photoreceptor degeneration model. Representative retinal sections from TRACR infected Ai63 at P5 and treated with MNU or vehicle at P20, then harvested at P38. Sections are stained for Sender (GFP, green), Receiver (Myc, white), Reporter (TdTomato, magenta), and nuclei (Hoechst, blue). Robust reporter activation is observed in vehicle-treated controls but is strongly reduced in MNU-treated eyes. Right: Percentage of Reporter-positive Receiver-expressing cells in control and MNU-treated mice. N = 3 sections per retina, 3 retinas per condition. Data are shown as mean ± SD. p =0.0005, Welch's t-test. Scale bar, 20 μm (B,C). .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 8, 2026. ; https://doi.org/10.64898/2026.02.08.704659doi: bioRxiv preprint IPL B IPL ONL OPL INL C A Hoechst Sender Receiver Reporter Hoechst Sender Receiver Reporter Control Rod TeNT INL OPL Rod PhR: TeNT; Ai63 Hoechst SCGN Reporter ONL OPL INL IPL %Marker+ Reporter+ cells %Reporter+ Rec cells D Control Rod ELFN1 cKO Hoechst Sender Receiver Reporter Hoechst Sender Receiver Reporter E Hoechst SCGN Reporter %Reporter+ Rec cells%Marker+ Reporter+ cells TeNTCont Elfn cKOCont ONL ONL OPL INL IPL ONL OPL INL IPL ONL OPL INL IPL ControlRod TeNT Rod ELFN1 cKO RBPCBP RBPCBP 1. 1. TeNTElfn1 cKoELFN1 X X X PhR Sender ON BP Receiver SV release X X *** *** **** ** Elfn1 cKO; Ai63 Figure 6 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 8, 2026. ; https://doi.org/10.64898/2026.02.08.704659doi: bioRxiv preprint Figure 6. TRACR activation requires intact photoreceptor-to-bipolar cell synapses. (A) Schematic of experimental strategies to block photoreceptor neurotransmission by rod-specific expression of tetanus toxin (Nrl-Cre;Rosa-TeNT), and disrupt transsynaptic connectivity rod-specific deletion of Elfn1 (Nrl-Cre;Elfn1flox). Mice were crossed with the Ai63 reporter line, infected with AAV-PhR-Sender and AAV-BP-Receiver at P5 and analyzed at P20. (B) TRACR activation is reduced when neurotransmitter release from rods is blocked. Representative P20 retinal cross-sections from control (Rosa-TeNT;Ai63) and Nrl-Cre;Rosa-TeNT;Ai63 mice infected with Sender and Receiver AAVs and stained for Sender (GFP, green), Receiver (Myc, white), Reporter (TdTomato, magenta), and nuclei (Hoechst, blue). Right: Percentage of Reporter-positive Receiver-expressing cells. N = 3 sections per retina, 3 retinas per condition. Data are shown as mean ± SD. ***p = 0.0003, Welch's t-test. (C) TRACR activation is preserved in cone bipolar cells when rod neurotransmitter release is blocked. Representative retinal cross-sections from Nrl-Cre; Rosa-TeNT;Ai63 mice stained for Reporter (TdTomato, magenta) and cone bipolar cells (SCGN, cyan). Right: Percentage of cone bipolar cells (SCGN-positive) or rod bipolar cells (PKC-positive) Reporter-positive cells. Data are shown as mean ± SD. ***p = 0.0004, Welch's t-test. (D) TRACR activation is reduced when rod–bipolar synapse formation is disrupted. Representative P20 retinal sections from control (Elfn1flox;Ai63) and Nrl-Cre;Elfn1flox;Ai63 mice infected with Sender and Receiver AAVs and stained for Sender (GFP, green), Receiver (Myc, white), Reporter (TdTomato, magenta), and nuclei (Hoechst, blue). Right: Percentage of reporter-positive Receiver-expressing cells. N= 3 sections per retina, 3 retinas (control) and 7 retinas (Nrl-Cre;Elfn1flox;Ai63). Data are shown as mean ± SD. ****p < 0.0001. (E) TRACR activation is preserved in cone bipolar cells when rod–bipolar connectivity is disrupted. Representative retinal sections from Nrl-Cre;Elfn1flox;Ai63 mice stained for Reporter (TdTomato, magenta) and cone bipolar cells (SCGN, cyan). Right: Percentage of cone bipolar cells (SCGN-positive) or rod bipolar cells (PKC-positive) Reporter-positive cells. N=3 sections per retina, 7 retinas per SCGN marker, 5 per PKC marker. Data are shown as mean ± SD. **p = 0.0017. Scale bar, 20 μm (B-E). .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted February 8, 2026. ; https://doi.org/10.64898/2026.02.08.704659doi: bioRxiv preprint