All-optical voltage interrogation for probing synaptic plasticity in vivo

The study addresses how to directly measure synaptic efficacy and the induction rules of synaptic plasticity at identified synapses in vivo, a prerequisite for understanding learning and memory. Traditional intracellular recordings are brief and low-yield in intact brains, leading researchers to use calcium imaging as a proxy. Recent two-photon-suitable genetically encoded voltage indicators (GEVIs) enable direct measurement of membrane voltage in identified neuronal populations in vivo. The authors aim to develop an all-optical methodology that selectively drives presynaptic inputs and optically reads postsynaptic voltage with high spatiotemporal resolution over long times in awake mice. They focus on cerebellar Purkinje cells, leveraging optogenetic activation of granule cell inputs and sensory-evoked climbing fiber inputs, to quantify subthreshold (IPSPs/EPSPs) and suprathreshold (complex spikes) signals and to probe synaptic plasticity induction rules.

The paper situates the work within: (1) challenges of direct intracellular recordings in vivo and common reliance on two-photon calcium imaging as a proxy for spiking and synaptic activation; (2) emergence of two-photon-compatible GEVIs enabling direct voltage readout in vivo; (3) cerebellar microcircuit principles, including microzones driven by coupled inferior olive neurons and known synchrony of Purkinje cell complex spikes; (4) classical cerebellar plasticity paradigms involving conjunctive parallel fiber (granule cell) and climbing fiber activation; and (5) growing evidence that inhibitory synapses exhibit plasticity. This context motivates an all-optical approach combining GEVI-based voltage imaging with optogenetics and sensory stimulation to interrogate identified synapses and their plasticity in awake animals.

- GEVI development: JEDI-2P was modified by inserting a tryptophan between cpGFP and the voltage-sensing domain, yielding JEDI-2Psub with enhanced subthreshold sensitivity and photostability. GEVI performance was characterized in HEK293A cells under voltage clamp using two-photon excitation (940 nm), assessing responses to single spikes, spike trains, bursts on subthreshold depolarizations, and voltage steps from -120 to +50 mV. Brightness and photostability were also benchmarked with a high-throughput two-photon screening platform. - Animal model and expression strategy: Math1-Cre mice (3–5 months) were used. AAVs were co-injected into cerebellar vermis lobules V/VI: pAAV-CaMKIIa-JEDI-2Psub-Kv-WPRE (AAV2/1) for selective Purkinje cell expression; Cre-dependent ChRmine-mScarlet for granule cell expression (ssAAV-9/2-hEF1a-dlox-ChRmine_MRS_mScarlet_ERES_WPRE-hGHP(A)). Headplate and cranial window implantation were performed in a single surgery. For pharmacology, coverslips with an access port enabled topical application. - In vivo imaging and stimulation: Awake head-fixed mice on a running wheel were imaged with resonant-scanning two-photon microscopy at 440 Hz (typical FOV 208 × 24 µm; depth 50–100 µm). JEDI-2Psub was excited at 940 nm. Granule cells were optogenetically stimulated with a 590 nm LED (2 ms pulses; typical intensity ~8 mW/mm^2). Sensory stimulation (25 ms air puff to the whisker pad) evoked climbing fiber input. To protect detectors during LED pulses, electronically gated PMTs introduced a 5–8 ms deadtime after LED onset. - Signal types recorded: Purkinje cell dendritic voltage signals included spontaneous and sensory-evoked complex spikes (CS), optogenetically evoked inhibitory postsynaptic potentials (IPSPs), and occasional EPSPs or dendritic spikes after GrC activation. - Data processing: Motion correction (Suite2p), manual dendrite ROI identification, intensity-weighted skeletonization and segmentation into ~4.8–5 µm segments. ΔF/F0 computed with correction for photobleaching and LED gating intervals. Baseline extracted via low-pass filtering; spikes detected as >3 SD above baseline. Triggered-average upsampling used hardware triggers to improve temporal precision for stimulation and spikes. - Spatiotemporal analyses: Cross-correlation of CS timing across neighboring dendrites to identify microzone synchrony; analysis of correlations of IPSP amplitudes/latencies across neighboring Purkinje cells; spatial heterogeneity quantified via coefficient of variation (CV) across dendritic segments for CS and IPSPs. - Pharmacology: Topical gabazine (200 µM) applied through cranial window port to confirm GABAA-mediated IPSPs by comparing optogenetically evoked responses pre/post-application. - Plasticity induction protocol: Conjunctive pairing protocol in vivo where sensory-evoked CF input preceded optogenetic GrC activation by 150 ms; 300 pairings at 1 Hz. IPSP amplitudes were measured before pairing (2 ms LED, 0.5 Hz × 60 trials) and 40 min after pairing across multiple FOVs. Controls included omission of pairing and reversal of stimulus order (GrC before CF). Cell identity across sessions was matched by morphology, correlation matrices of segments, CS shape, and firing rate. Statistics included Wilcoxon tests, linear regression with Wald tests, and linear mixed-effect models with animal and FOV as random effects.

- GEVI engineering: JEDI-2Psub displayed larger single-spike responses than JEDI-2P under two-photon excitation (−ΔF/F0: −34.1 ± 6.8% vs −23.4 ± 3.5% in HEK293A cells), improved photostability, and a left-shifted voltage sensitivity yielding ~3.5× larger responses near resting potentials. - Robust in vivo dendritic voltage readout: Spontaneous Purkinje cell dendritic complex spikes (CS) were prominent (−ΔF/F0 = −31.2 ± 4.4%, FWHM 10.5 ± 1.8 ms, rate 1.3 ± 0.4 Hz; discriminability index d′ = 5.9 ± 1.3). CS amplitude increased with baseline depolarization. - Sensory-evoked responses: Whisker air puffs evoked diverse dendritic signals including inhibitory and excitatory responses. Sensory-evoked CS waveforms matched spontaneous CS, with a mean time to peak of 65.1 ± 13.1 ms (n = 43 cells), confirming CF-driven excitation. - Optogenetically evoked IPSPs: GrC activation produced graded hyperpolarizations in PCs scaling with optical intensity, with onset 9.6 ± 4.8 ms and time-to-peak 85.4 ± 24.4 ms. Gabazine significantly reduced IPSP amplitudes (control ΔF/F0 = 4.1 ± 2.8% to 1.6 ± 1.6%; Wilcoxon signed-rank p = 1.26 × 10^-4), confirming GABAA mediation. Occasionally, EPSPs/dendritic spikes were observed (−ΔF/F0 = −10.5 ± 2.9%, peak 10.3 ± 0.6 ms), though rare due to feed-forward inhibition and 5–8 ms post-LED deadtime. - Microzone synchrony: Neighboring PCs exhibited precise synchrony of spontaneous CS (cross-correlogram peak width ~6.4 ms). Sensory-evoked CS peak latency and probability were correlated between neighbors (R = 0.490, p = 1.15 × 10^-3; R = 0.590, p = 4.85 × 10^-5), whereas CS amplitudes were not (R ≈ 0.22–0.29, ns). - Shared inhibition across neighbors: IPSP amplitudes and latencies were highly correlated across neighboring PCs: sensory-evoked IPSP amplitudes R = 0.843 (p = 4.55 × 10^-12), optically evoked IPSP amplitudes R = 0.807 (p = 1.90 × 10^-10); peak latencies R = 0.500 (p = 8.71 × 10^-4) and R = 0.474 (p = 1.74 × 10^-3), respectively, suggesting shared presynaptic interneurons. - Spatial distribution across dendrites: CS were uniformly distributed across the dendritic tree (CV ~0.079 ± 0.032 for sensory-evoked; 0.095 ± 0.043 for spontaneous), consistent with global CF depolarization. IPSPs were more heterogeneous (CV = 0.130 ± 0.061), reflecting discrete inhibitory synapse activation. - Plasticity induction (all-optical): Pairing CF input followed 150 ms later by GrC activation (300 pairings @1 Hz) induced robust LTP of IPSPs that persisted for at least 40 min (mean IPSP ΔF/F0: 7.9 ± 4.1% pre vs 9.9 ± 4.6% post; Wilcoxon signed-rank p = 8.80 × 10^-3, n = 32 cells, N = 4 mice). No potentiation in controls omitting pairing (LME: p = 4.92 × 10^-4), nor with reversed order (GrC→CF). - Determinants and network effects of LTP: Smaller baseline IPSPs predicted larger potentiation (R = −0.543, p = 1.32 × 10^-3). LTP magnitude was correlated across neighboring PCs (R = 0.611, p = 1.95 × 10^-3), increasing intercell IPSP correlations post-pairing. Within dendrites, LTP reduced spatial variability of IPSP amplitudes (CV decrease vs control; Mann–Whitney p = 3.94 × 10^-3), while CS spatial distribution remained unchanged (p = 0.204).

Combining two-photon voltage imaging with optogenetic and sensory stimulation, the study demonstrates all-optical induction and readout of synaptic plasticity at identified cerebellar synapses in awake mice. JEDI-2Psub’s enhanced sensitivity around resting potentials enabled reliable detection of subthreshold IPSPs alongside suprathreshold complex spikes over extended periods. The approach revealed precise microzone synchrony of CF-driven complex spikes and strong shared inhibitory structure across neighboring Purkinje cells. Critically, conjunctive CF→GrC pairing produced long-term potentiation of inhibitory inputs, indicating inhibitory synapses as significant, plastic substrates in cerebellar circuits. Such inhibitory LTP could counterbalance excitatory LTP and, by regulating CF calcium signaling, constrain future plasticity induction. The method’s ability to directly quantify synaptic efficacy changes in vivo, across dendrites and neighboring neurons, provides a powerful framework to dissect the rules of plasticity and their network-level consequences during behavior.

This work introduces an all-optical strategy for probing synaptic plasticity in vivo by integrating a subthreshold-optimized GEVI (JEDI-2Psub), two-photon voltage imaging, optogenetic activation of granule cells, and sensory-evoked climbing fiber inputs in awake mice. Key contributions include: (1) engineering JEDI-2Psub for enhanced subthreshold sensitivity and stability; (2) robust mapping of subthreshold and suprathreshold dendritic signals across neighboring Purkinje cells; and (3) demonstration of timing-dependent LTP at inhibitory synapses in vivo, with coordinated potentiation across neighboring cells and normalization of dendritic inhibitory responses. Future directions include scaling to faster imaging modalities, monitoring synaptic efficacy across full dendritic trees and individual spines, extending to behavioral tasks, and elucidating cellular mechanisms and timing rules underlying inhibitory plasticity to link synaptic changes to learning.

- Experimental design: Experiments were not randomized; investigators were not blinded during data acquisition (though blinded to timing condition during analysis). Sample sizes were not predetermined but are comparable to prior in vivo two-photon voltage imaging studies. - Selection and generalizability: PCs not responding to optogenetic activation were not recorded, introducing selection bias. Findings are currently limited to cerebellar Purkinje cells in Math1-Cre mice. - Measurement constraints: An electronically gated PMT introduced a 5–8 ms deadtime after LED onset, likely reducing detection of early EPSPs; feed-forward inhibition further masked EPSPs. Imaging sessions were limited (~3 min per session) to mitigate photobleaching. - Duration and scope of plasticity: LTP of IPSPs was tracked for 40 min post-pairing; longer-term stability and underlying mechanisms were not addressed. - Potential confounds: In control datasets, slight IPSP depression was observed (presumably due to calcium influx from optogenetic activation).