An engineered channelrhodopsin optimized for axon terminal activation and circuit mapping

The study addresses a key limitation in optogenetics: widely used channelrhodopsin-2 (ChR2) variants lack selective enrichment at presynaptic axon terminals, leading to confounding activation of somata, dendrites, and passing axons during photostimulation. This limits precise activation of long-range projections for circuit mapping, antidromic identification, and selective synaptic release. The authors hypothesize that engineering a localization tag that routes ChR2 to presynaptic terminals while reducing somatodendritic expression will enable efficient terminal activation and cleaner circuit mapping. They propose using the C-terminal domain of metabotropic glutamate receptor 2 (mGluR2), which is enriched presynaptically, and augmenting it with a proteolytic degradation motif (PEST) and an axon-targeting element (ATE) to minimize somatodendritic accumulation and enhance distal trafficking. The goals are to demonstrate terminal-enriched localization in vitro and in vivo, validate preserved synaptic physiology, improve terminal activation efficiency at lower light, suppress somatic excitability, and enhance the reliability and specificity of spike collision tests in vivo.

Prior work has improved ChR2 performance by subcellular targeting motifs: Golgi trafficking signals increased surface expression and photocurrents; somatodendritic enrichment achieved via myosin V binding domains, KxE motifs, or neuroligin-derived sequences; postsynaptic density targeting via ETQV PDZ-binding motif. Attempts to localize ChR2 to the AIS using ankyrin G-binding domains disrupted endogenous Nav clusters and impaired spiking. Myosin Vb binding domain enhanced local axonal activation near soma but not long-range terminals. Thus, tools that efficiently and selectively target ChR2 to distal axons and presynaptic terminals remain needed. mGluR2 is known to localize presynaptically, but reports differ on its axonal transport depending on developmental stage and expression system. The authors leverage mGluR2’s C-terminus for presynaptic enrichment and add PEST (to reduce somatodendritic levels) and ATE (to promote axonal targeting), aiming to overcome prior limitations while preserving normal release properties.

Design and molecular engineering: ChR2(134)-YFP was used as the optogenetic actuator. Presynaptic-localizing tags were fused C-terminally via a modified multi-cloning site: (i) CAST-interacting region, (ii) RIM1 N-terminal region, (iii) myosin V binding domain (MYDB/MViBD), and (iv) mGluR2 C-terminal domain. To suppress somatodendritic accumulation and enhance axonal targeting, PEST (proteolytic) and ATE (axon-targeting element) sequences were added after ChR2-YFP-mGluR2, generating mGluR2-PA. Constructs were packaged in AAV vectors under the human synapsin 1 promoter. In vitro localization assays: Primary cultured hippocampal neurons were transduced with AAVs. Expression was assessed by Western blot and immunocytochemistry. Subcellular localization was quantified using immunostaining for presynaptic marker Bassoon, axonal marker Tau, and dendritic marker MAP2. Colocalization was quantified with Pearson’s R. Aggregation and puncta were visually assessed. In vivo anatomical localization: AAVs expressing control ChR2-YFP or mGluR2-PA-tagged ChR2-YFP were injected into hippocampal CA3 and cortical regions in adult mice. Projection labeling in contralateral hippocampus/cortex was imaged; projection-site fluorescence intensity was normalized to injection-site intensity. Pre-embedding immunofluorescence and electron microscopy were used to classify YFP-positive profiles into axons vs terminals and assess distribution at active zones (AZ) vs non-AZ within terminals. Physiological validation in long-range pathway: Glutamatergic neurons in parabrachial nucleus (PB) were transduced, and optogenetically evoked synaptic responses were recorded in central amygdala (CeA) neurons in acute slices. Light-evoked EPSCs (leEPSCs) were recorded across tag variants, including control, mGluR2-PA, mGluR2, MYDB/MViBD, CAST, and RIM1 (RIM1 showed poor expression). Expression levels were titer-adjusted to compare function. Kinetics (latency, jitter, rise/decay), success rates, amplitudes, and paired-pulse ratio (PPR) were measured. Tetrodotoxin (TTX, 1 μM) tested action potential dependence. Input-output comparisons for terminal vs soma activation: In projection sites, graded light intensities were applied to measure EPSC amplitudes; at injection sites (PB), whole-cell current-clamp recordings from YFP+ neurons assessed spiking probability across light intensities, comparing control vs mGluR2-PA. In vivo spike collision test: In rats, AAVs were injected into left M1; antidromic optical stimulation was applied to right M1 while recording spikes in left M1 with silicon probes. Spike collision success (disappearance of evoked antidromic spikes when preceded by spontaneous spikes) was quantified across sessions. Multiple unit activity (MUA) after photostimulation was analyzed to assess polysynaptic/noise activation. Latency and jitter of antidromic spikes were measured. Statistics: Data are mean ± SEM. Tests included one-way ANOVA with post hoc Tukey or Bonferroni, Student’s t-tests, two-way repeated-measures ANOVA with simple effects tests, Kruskal–Wallis with Steel–Dwass, χ² tests. Significance defined as P < 0.05.

- Screening of presynaptic tags: CAST, mGluR2, and MYDB maintained ChR2-YFP expression; RIM1 greatly attenuated expression. CAST and MYDB exhibited nonspecific aggregation; mGluR2 produced punctate signals colocalized with Bassoon. - Adding PEST and ATE to mGluR2 (mGluR2-PA) significantly reduced overall expression compared to control and mGluR2 alone (one-way ANOVA, P < 0.01) and decreased somatodendritic signals while preserving axonal/terminal localization. - Colocalization (Pearson’s R): mGluR2 increased colocalization with Bassoon vs control; mGluR2-PA maintained presynaptic colocalization similar to control while greatly reducing dendritic colocalization with MAP2; Tau (axonal) colocalization was preserved. Sample sizes: Bassoon control n=26, mGluR2 n=118, mGluR2-PA n=17; Tau/MAP2 control n=28, mGluR2 n=17, mGluR2-PA n=29. Significant effects: *P < 0.05, **P < 0.001. - In vivo hippocampal projections: Normalized contralateral projection-site fluorescence was significantly higher with mGluR2-PA vs control (Student’s t-test, *P < 0.05, **P < 0.01), indicating enhanced accumulation at terminals. - EM classification (Fig. 2d): mGluR2-PA significantly increased the proportion of YFP signals at terminals vs axons (χ² test, terminal vs axon counts: control terminal 424, axon 297; mGluR2-PA terminal 389, axon 140; P = 0.0006). Within terminals, distribution at AZ vs non-AZ was not significantly different between groups. - Functional validation in PB→CeA pathway: Across control, mGluR2-PA, mGluR2, MViBD, and CAST (RIM1 excluded), leEPSC amplitudes were comparable on average; large leEPSCs were more frequently observed in mGluR2-PA. Kinetic parameters (latency, jitter, rise, decay) showed no significant differences. PPR showed no significant differences, indicating preserved presynaptic release probability. TTX (1 μM) abolished leEPSCs in all groups, confirming AP-dependent synaptic transmission. - Terminal activation efficiency (Fig. 4): Input–output curves for leEPSCs showed significantly greater responses at lower light intensities in mGluR2-PA vs control (two-way RM ANOVA interaction P = 0.018), e.g., responses maintained at 6 mW in mGluR2-PA while control dropped markedly at 2.6–2.4 mW. - Soma/dendrite excitability (Fig. 4): Spiking probability in PB somata required higher light intensities in mGluR2-PA compared to control; input–output curve right-shifted (two-way RM ANOVA interaction P = 0.0139). Probabilities at 0.6 and 0.24 mW were significantly lower in mGluR2-PA. - In vivo spike collision test (Fig. 5): Successful collision identifications: control 7/16 sessions (n=4 rats), mGluR2-PA 12/16 sessions (n=3 rats). MUA increases after single photostimuli were significantly reduced in mGluR2-PA vs control during the first 100 ms post-stimulation, indicating less polysynaptic/noise activation. Antidromic spike latency and jitter were not significantly different (control: latency 12.1 ± 1.7 ms, jitter 2.08 ± 0.3 ms; mGluR2-PA: latency 13.1 ± 2.5 ms, jitter 0.28 ± 0.07 ms).

The engineered mGluR2-PA tag effectively routes ChR2 to presynaptic terminals while suppressing somatic and dendritic expression, directly addressing the need for selective activation of long-range axons. EM and fluorescence analyses confirmed increased terminal accumulation without altering distribution at active zones. Physiological assessments showed that terminal photostimulation with mGluR2-PA evokes reliable, AP-dependent synaptic transmission at lower light intensities than control, while somatic spiking requires stronger light, effectively biasing activation toward terminals. Importantly, PPR and leEPSC kinetics were unaffected, indicating that the tag did not distort presynaptic release machinery. In vivo spike collision tests benefited from reduced polysynaptic and off-target activation (lower MUA noise) without compromising antidromic spike fidelity, improving spatial and temporal resolution for mapping axonal projections. These findings validate a general strategy for enhancing the specificity of optogenetic circuit mapping by presynaptic targeting, with potential applicability to other opsins and neural systems. Developmental and context-dependent transport mechanisms of mGluR2 may explain discrepancies with prior reports, and removal of PDZ-binding motifs alongside inclusion of ATE likely contributed to presynaptic enrichment observed here.

This work introduces mGluR2-PA, a presynaptic localization tag that, when fused to ChR2-YFP, enriches the actuator at axon terminals, reduces somatodendritic expression, and improves the efficiency and specificity of terminal activation. The approach enables lower-intensity photostimulation to drive synaptic release, preserves endogenous presynaptic function, and enhances in vivo antidromic identification with reduced polysynaptic noise. The tag provides a valuable tool for dissecting long-range circuit connectivity and synaptic mechanisms in vitro and in vivo. Future directions include optimizing tags for even more selective terminal-only activation, testing across diverse cell types and circuits, adapting the strategy to other optogenetic actuators and indicators, and implementing in transgenic models for uniform expression with minimal off-target activation.

- Despite reduced somatodendritic expression, mGluR2-PA does not entirely eliminate somatic responsiveness; strong light can still evoke spikes at the soma/dendrites. - Subthreshold somatic stimulation that failed to evoke action potentials also did not trigger synaptic release at terminals under tested conditions, indicating incomplete decoupling of terminal activation from somatic excitability. - EM analyses found no change in AZ vs non-AZ distribution, suggesting terminals are enriched broadly rather than precisely at release sites; finer AZ targeting may be needed for some applications. - Discrepancies with prior reports on mGluR2 axonal transport likely reflect developmental stage and expression system differences; generalizability across developmental stages and species requires further study. - AAV-based expression can be heterogeneous; although ectopic expression was minimal here, the extent to which mGluR2-PA affects off-target expression after viral transduction remains to be fully evaluated. - Some alternative presynaptic tags (e.g., CAST, MYDB) showed aggregation or poor expression (RIM1), limiting their utility and highlighting the need for expression-level control.