A fluorescent sensor for spatiotemporally resolved imaging of endocannabinoid dynamics in vivo

Abstract

Endocannabinoids (eCBs) are retrograde neuromodulators with important functions in a wide range of physiological processes, but their in vivo dynamics remain largely uncharacterized. Here, we developed a genetically encoded eCB sensor called GRAB_eCB2.0. GRAB_eCB2.0 consists of a circular-permutated EGFP and the human CB1 cannabinoid receptor, providing cell membrane trafficking, second-resolution kinetics, high specificity for eCBs, and shows a robust fluorescence response at physiological eCB concentrations. Using GRAB_eCB2.0, we monitored evoked and spontaneous changes in eCB dynamics in cultured neurons and acute brain slices. We observed spontaneous compartmentalized eCB transients in cultured neurons, and eCB transients from single axonal boutons in acute brain slices, suggesting constrained, localized eCB signaling. Expressing GRAB_eCB2.0 in the mouse brain, we observed foot shock-elicited and running-triggered eCB signaling in the basolateral amygdala and hippocampus, respectively. In a mouse model of epilepsy, we observed a spreading wave of eCB release that followed a Ca²⁺ wave through the hippocampus. GRAB_eCB2.0 is a robust probe for eCB release in vivo.

Cannabis derivatives have long been used for medicinal and recreational purposes across cultures¹. Bioactive compounds in cannabis, phytocannabinoids, exert their function by “hijacking” the body’s endogenous cannabinoid (endocannabinoid, or eCB) system. The biological function of eCBs—mainly two lipid metabolites 2-arachidonoylglycerol (2-AG) and anandamide (AEA)—is primarily mediated by the activation of type1 and type 2 cannabinoid receptors (CB1R and CB2R)². eCBs are widely distributed throughout the peripheral and central nervous system, where they serve as important neuromodulators. Interestingly, unlike other classical neurotransmitters stored in synaptic vesicles and released from the presynaptic terminal, eCBs are typically produced and released from the postsynaptic compartment in a neuronal activity-dependent manner, then retrogradely travel to the presynaptic terminal and activate the CB1R, activation of which often results in an inhibition of presynaptic neurotransmitter release^3,4. In addition, eCBs also play a role in glial cells and in intracellular organelles^5-9. In the brain, eCBs participate in the short-term and long-term synaptic plasticity of glutamatergic and gamma-aminobutyric acid (GABA)-ergic synapses in a variety of regions, including the cerebral cortex, hippocampus, striatum, ventral tegmental area, amygdala and cerebellum^4,10, playing important roles in a wide range of physiological processes such as development, emotional state, pain, the sleep/wake cycle, energy metabolism, reward, and learning and memory^11-15. Given the broad distribution and variety of functions of eCBs, dysregulation of the eCB system has been associated with a plethora of disorders, including neuropsychiatric and neurodegenerative diseases, epilepsy, cancer, and others^16-18. The eCB system has therefore emerged as a promising target for treating neurological diseases^19,20.

Although we know much about the eCB biochemistry and physiology, the spatiotemporal dynamics of eCB release in the brain remains largely unknown. Signaling via endocannabinoid receptors is believed to last on the order of seconds and over a distance on the order of tens of microns²¹. However, this assumption has not been tested definitively, largely because existing methods for measuring eCB signaling lack the necessary spatiotemporal resolution. For example, although qualitative and quantitative measurement of eCBs in brain tissues can provide valuable information regarding eCB levels, these measurements usually require the extraction, purification and analysis of lipids by chromatography and mass spectrometry^22,23, therefore, this approach has poor spatial and temporal resolution and cannot be used to measure eCBs in vivo. Another approach is electrophysiology coupled with pharmacology and/or genetics, which is often used to indirectly measure eCB activity by measuring eCB-mediated synaptic modulation^21,24-26; however, this method is mostly used in in vitro preparations and has relatively low spatial resolution. Microdialysis, while challenging for hydrophobic lipid molecules, has been used to monitor eCB abundance in the brain during pharmacological manipulations and behaviors^27,28, but it has a long sampling interval (at least 5 minutes) that is well beyond the time scale of synaptic plasticity mediated by eCBs (~sub-second to seconds), preventing the accurate detection of eCBs in real time in vivo. Therefore, development of an in vivo eCB detection tool with satisfactory spatiotemporal resolution would meet a clear need in this field.

Recently, we and others developed a series of genetically-encoded tools for sensing neurotransmitters and neuromodulators based on G protein-coupled receptors (GPCRs) and circular-permutated (cp) fluorescent proteins^29-37. Using this highly successful strategy, we developed a novel GPCR activation-based (GRAB) eCB sensor called GRAB_eCB2.0 (or simplified as eCB2.0) based on the human CB1R and cpEGFP. The eCB2.0 sensor has high specificity for eCBs, kinetics on the order seconds, and a fluorescence response of approximately 950% to 2-AG and 500% to AEA, respectively. After validating the in vitro performance of eCB2.0 in both cultured cells and acute brain slices, we then expressed the sensor in mice and reliably monitored foot-shock evoked eCB signals in the basolateral amygdala in freely moving mice and eCB dynamics in the mouse hippocampus during running and seizure activity.

RESULTS

Development and in vitro characterization of GRAB_eCB sensors

Among the two eCB receptors, we chose CB1R as the scaffold for developing a GRAB eCB sensor, because this receptor has a higher affinity for eCBs than CB2R³⁸. We first inserted the intracellular loop 3 (ICL3)-cpEGFP module of our recently developed GRAB_NE sensor³² into the corresponding ICL3 in the human CB1R (Fig. 1a). After screening several insertion sites, we generated the prototype eCB sensor named GRAB_eCB0.1 (eCB0.1) with 30% fluorescent response to 10 μM 2-AG (Extended Data Fig. 1a). We speculated that the long ICL3 from GRAB_NE may resist the conformational coupling between CB1R and cpEGFP, therefore, we truncated the ICL3 of eCB0.1 and generated GRAB_eCB1.0 (eCB1.0) with 100% response to 2-AG (Fig 1b and Extended Data Fig. 1a). To improve the dynamic range of our eCB sensor, we then selected 8 residues in cpEGFP for individual randomized mutation based our experience gained through the development of previous GRAB sensors^{29,31-33,35-37} (Extended Data Fig. 1a). Combining several single-mutation candidates—each with improved performance—resulted in the GRAB_eCB1.5 sensor (eCB1.5), which has a 2-fold higher response than eCB1.0 (Fig. 1b and Extended Data Fig. 1a). We next focused on the receptor’s ligand binding pocket in order to further improve the sensor’s dynamic range and affinity. The residues F177^2.64, V196^3.32 and S383^7.39 were selected for targeted screening based on the studies of CB1R structure^39-44 (Extended Data Fig. 1a). Interestingly, we found that introducing the S383^7.39T mutation in eCB1.5 produced an increased response to 2-AG, whereas adding the F177^2.64A mutation abolished the response to 2-AG (Extended Data Fig. 1a). We therefore selected the eCB1.5 with the S383^7.39T mutation as the second-generation GRAB_eCB2.0 sensor (eCB2.0), and eCB1.5 with both the S383^7.39T and F177^2.64A mutations as a non-responsive negative control, which we call GRAB_eCBmut sensor (eCBmut) (Fig. 1b and Extended Data Fig. 1b,c).

Fig. 1 ∣. Development, optimization, and characterization of GRAB_eCB sensors in HEK293T cells.

a, Schematic diagram depicting the design and principle of the GRAB_eCB sensor, consisting of the CB1 receptor and circular-permutated GFP. Ligand binding activates the sensor, inducing a change in fluorescence.

b, Screening and optimization steps of GRAB_eCB sensors and the normalized fluorescence response to 10 μM 2-AG.

c, Expression and fluorescence change in response to 100 μM 2-AG and AEA in HEK293T cells expressing eCB2.0. Similar results were observed for more than 20 cells. Scale bar, 30 μm.

d, Dose-response curves measured in HEK293T cells expressing eCB2.0 or eCBmut, with the corresponding EC₅₀ values for 2-AG and AEA shown; n = 6 experiments, mean ± s.e.m.

e, Normalized fluorescence change in response to the indicated compounds (each at 10 μM) measured in cells expressing eCB2.0. Where indicated, the CB1R inverse agonist AM251 was also added. LPA, lysophosphatidic acid; S1P, sphingosine-1-phosphate; ACh, acetylcholine; DA, dopamine; GABA, gamma-aminobutyric acid; Glu, glutamate; Gly, glycine; NE, norepinephrine; 5-HT, 5-hydroxytryptamine; His, histamine; Epi, epinephrine; Ado, adenosine; Tyr, tyramine. n = 3 wells for ACh, Glu, Gly, NE, 5-HT, His, Epi, Ado and Tyr; n = 4 for other groups, mean ± s.e.m. One-way ANOVA test was performed for all groups: P=8.00E-26; Tukey tests were performed post hoc: P=0 (between 2-AG and AEA+AM251, 2-AG+AM251, LPA, …, or Tyr), P=0 (between AEA and AEA+AM251, 2-AG+AM251, LPA, …, or Tyr).

f, Illustration of the localized puffing system using a glass pipette containing 100 μM 2-AG and/or AM251 positioned above an eCB2.0-expressing cell. The dotted black line indicates the region of interest for line scanning. Similar results were observed for more than 10 cells. Scale bar, 30 μm.

g, Change in eCB2.0 fluorescence was measured in an eCB2.0-expressing cell using line scanning; where indicated, 2-AG and AM251 were puffed on the cell. The graph at the right summarizes the on and off time constants measured upon application of 2-AG and upon application of AM251, respectively; n = 11 (τ_on) and 4 (τ_off) cells, mean ± s.e.m.

h, One-photon (1P) excitation and emission spectra and two-photon (2P) excitation spectra of eCB2.0 in the absence and presence of 2-AG. Excitation and emission peaks were labeled.

***, p < 0.001.

When expressed in HEK293T cells, both the eCB2.0 and eCBmut sensors trafficked to the cell membrane (Fig. 1c). Upon ligand application, eCB2.0 had a concentration-dependent fluorescence increases to both 2-AG and AEA, with a maximum response of approximately 3 fold relative to baseline and the half maximal effective concentrations (EC₅₀) for 2-AG and AEA of 3.1 μM and 0.2 μM, respectively; in contrast, eCBmut showed no response to 2-AG or AEA at all concentrations tested (Fig. 1d). We then tested whether the sensor’s response is specific to eCBs compared to other neurotransmitters. We found that eCB2.0 responded robustly to both 10 μM AEA and 2-AG, and the response was abolished by the CB1R inverse agonist AM251; moreover, no other neurotransmitters or neuromodulators tested elicited any response in cells expressing eCB2.0 (Fig. 1e). In addition, eCB2.0 showed robust response to synthetic CB1R agonists WIN55212-2 and CP55940, but weak response to CB1R partial agonist Δ−9-THC (Extended Data Fig. 2).

Next, we measured the kinetics of eCB2.0 signaling using a rapid localized solution application system in which compounds were puffed directly on the cell (Fig. 1f). To measure the onset rate (τ_on), 100 μM 2-AG was puffed on eCB2.0 expressing cell; to measure the offset rate (τ_off), 100 μM AM251 was puffed in the presence of 10 μM 2-AG. Using this approach, we measured averaged τ_on and τ_off values of 1.6 s and 11.2 s, respectively (Fig. 1g). To characterize the photophysical properties of eCB2.0, we measured the fluorescent spectra and photostability of eCB2.0 under 1-photon (1P) and 2-photon (2P) conditions. The peak 1P excitation (500 nm) and emission (520 nm) as well as the peak 2P excitation (930 nm) wavelength of eCB2.0 were comparable to that of other common green fluorescent probes (Fig. 1h). Besides, eCB2.0 showed better photostability than EGFP under strong 1P illumination (Extended Data Fig. 3a-c) and better or comparable photostability than EGFP under 2P illumination (Extended Data Fig. 3d,e).

To examine whether eCB sensors couple with intracellular signaling pathways, we measured G-protein activation using two approaches. First, we used a chimeric Gα_q-i protein, which can be activated by G_i-coupled GPCR and can increase intracellular Ca²⁺ level through PLC and IP3. jRGECO1a was also used to monitor intracellular Ca²⁺ signals (Extended Data Fig. 3f). In wildtype CB1R expressing cells, 2-AG increased jRGECO1a fluorescent response in a dose-dependent manner, indicating 2-AG successfully activated CB1R and downstream G protein pathway; in contrast, 2-AG failed in inducing jRGECO1a response in eCB2.0 expressing cells, demonstrating no detectable coupling between eCB2.0 and G protein (Extended Data Fig. 3g,h). In addition, we found no significant difference of 2-AG induced jRGECO1a signals between CB1R expressing cells and CB1R + eCB2.0 co-expressing cells, suggesting the expression of eCB2.0 has minimal buffering effect on CB1R mediated signals (Extended Data Fig. 3g,h). Second, we used a bioluminescence resonance energy transfer (BRET) sensor to detect G_βγ activation. Treating cells expressing CB1R with 2-AG induced a robust increase in BRET, consistent with G protein activation; in contrast, 2-AG had no effect on BRET in mock-transfected control cells or in cells expressing either eCB2.0 or eCBmut (Extended Data Fig. 3i). In addition, we also measured β-arrestin recruitment using the Tango GPCR assay⁴⁵ and found that AEA induced a robust, concentration-dependent response in cells expressing CB1R but had no effect in control cells or cells expressing either eCB2.0 or eCBmut (Extended Data Fig. 3j). Lastly, we tested the eCB2.0 in HEK293T cells with or without co-expression of CB1R, and found that CB1R did not change the dose-dependent response of eCB2.0 to 2-AG (Extended Data Fig. 3k). Taken together, these data indicate that eCB2.0 sensor binds eCBs but does not couple to downstream effector proteins or affect CB1R mediated signals, and therefore likely does not affect cellular physiology; conversely, eCB2.0 is not significantly buffered by the CB1R expression either.

We then examined the expression pattern of the eCB sensor in neurons by sparsely expressing eCB2.0 in cultured rat cortical neurons using the calcium phosphate transfection. We found that eCB2.0 trafficked to the entire neuronal cell membrane, including the axons and dendrites, as shown by colocalization with the axonal presynaptic marker synaptophysin-mScarlet and the postsynaptic marker PSD95-mScarlet (Fig. 2a). To measure the response of eCB2.0 in neurons, we infected cultured rat cortical neurons using an adeno-associated virus (AAV) expressing either eCB2.0 or eCBmut under the control of the human SYN1 (synapsin) promoter to drive expression in all neurons (Fig. 2b). We found that both 2-AG and AEA elicited concentration-dependent fluorescence responses in neurons expressing eCB2.0, with a maximum fluorescence increase of 950% and 500%, and an EC₅₀ value of 9.0 μM and 0.8 μM, respectively; in contrast, neither 2-AG nor AEA elicited a response in neurons expressing eCBmut, even at 100 μM (Fig. 2b,c). We also found that eCB2.0 responses in neurites were higher than those in somata (Fig. 2d). Finally, bath application of the CB1R agonist WIN55212-2—which can activate eCB2.0 in HEK293T cells—to eCB2.0-expressing neurons induced a fluorescence increase that was stable for up to 2 hours and blocked completely by AM251 (Fig. 2e), suggesting that the sensor does not undergo arrestin-mediated internalization or desensitization and can be used for long-term monitoring of eCB activity.

Fig. 2 ∣. Characterization of GRAB_eCB sensors in primary cultured neurons.

a, Fluorescence microscopy images of primary cultured rat cortical neurons expressing eCB2.0 (green) and either synaptophysin-mScarlet (top row; red) or PSD95-mScarlet (bottom row; red). In the top row, arrows indicate axons; in the bottom row, arrowheads indicate dendrites and dendritic spines. Similar results were observed for more than 10 neurons. Scale bars, 30 μm (top row) and 15 μm (bottom row).

b, Fluorescence microscopy images and fluorescence response to 100 μM 2-AG (top row) or AEA (bottom row) in neurons expressing eCB2.0 (left) or eCBmut (right). The insets in the eCBmut images are contrast-enhanced to show expression of the sensor. Similar results were observed for more than 30 neurons. Scale bars, 30 μm.

c, (Left) example traces of ΔF/F₀ measured in an eCB2.0-expressing neuron; the indicated concentrations of 2-AG and AEA, followed by 100 μM AM251, were applied. (Right) dose-response curves measured in neurons expressing eCB2.0 or eCBmut, with the corresponding EC₅₀ values shown; n = 5 cultures each, mean ± s.e.m.

d, Summary of the change in eCB2.0 fluorescence in response to 100 μM 2-AG or AEA measured in the neurites and soma; n = 5 cultures each, mean ± s.e.m. Paired two-tailed Student’s t tests were performed: P=0.0073 (left) and 0.0068 (right).

e, Example images (left), trace (middle), and quantification (right) of the change in eCB2.0 fluorescence in response to a 2-hour application of WIN55212-2, followed by AM251; n = 3 cultures each, mean ± s.e.m. Similar results were observed for more than 20 neurons. Scale bar, 100 μm. One-way ANOVA test was performed all groups: P= 4.48E-13; Tukey tests were performed post hoc: P=0 (between Saline and 0.5 h, 1.0 h, 1.5 h or 2.0 h), P=0 (between AM251 and 0.5 h, 1.0 h, 1.5 h or 2.0 h).

***, p < 0.001; **, p < 0.01.

Imaging eCB in primary cultured neurons

Cultured neurons are commonly used for studying eCB mediated synaptic modulation²⁶. We therefore examined whether our eCB2.0 sensor can be used to detect the release of endogenous eCB in cultured rat cortical neurons expressing eCB2.0 together with a red glutamate sensor R^ncp-iGluSnFR⁴⁶. Applying electrical field stimuli (100 pulses at 50 Hz) elicited robust eCB and glutamate signals (Fig. 3a), demonstrating that eCB2.0 can reliably report endogenous eCB release and is compatible with red fluorescent indicators. Importantly, expression of eCB2.0 did not affect the electrically evoked R^ncp-iGluSnFR signals in neurons compared with that in control neurons, indicating eCB2.0 has minimal effect on neuronal glutamate release (Extended Data Fig. 4). We then expressed eCB2.0 in neurons loaded with a red fluorescent Ca²⁺ dye Calbryte-590 in order to simultaneously measure eCB release and changes in intracellular Ca²⁺. Field stimuli (100 pulses at 50 Hz) elicited robust responses with respect to both intracellular Ca²⁺ and eCB release (Fig. 3b). Moreover, the rise and decay kinetics of the calcium signal were faster than those of the eCB signal, consistent with the notion that eCB release requires neuronal activity⁴⁷. We also found a strong correlation between the peak Ca²⁺ signal and the peak eCB signal when applying increasing numbers of stimuli (R² = 0.99, Fig. 3c); importantly, in the absence of extracellular Ca²⁺, even 20 pulses were unable to elicit either a Ca²⁺ signal or an eCB2.0 response (Fig. 3c), confirming the requirement of calcium activity on eCB release.

Fig. 3 ∣. Release of endogenous eCB measured in primary cultured neurons.

a, Fluorescence microscopy images and fluorescence response measured in neurons co-expressing R^ncp-iGluSnFR (red) and eCB2.0 (green). Similar results were observed for more than 20 neurons. Scale bar, 200 μm.

b, Fluorescence microscopy images and fluorescence response measured in eCB2.0-expressing neurons preloaded with Calbryte-590 (red). Similar results were observed for more than 20 neurons. Scale bar, 200 μm.

c, Relative peak change in eCB2.0 fluorescence plotted against the relative peak change in Calbryte590 fluorescence measured in response to the indicated number of electrical pulses, normalized to the response evoked by 200 pulses; n = 4 cultures each, mean ± s.e.m. Also shown is the response to 20 electrical pulses with no extracellular Ca²⁺.

d, Diagram depicting the pathway for eCB synthesis. DAG, diacylglycerol; DAGL, diacylglycerol lipase; NAPE, N-arachidonoyl phosphatidylethanolamine; NAPE-PLD, NAPE-hydrolyzing phospholipase D.

e, Representative traces (left) and expanded traces (right) showing the change in eCB2.0 fluorescence in responses to 20 electrical pulses applied before (1) and after (2) DO34 application; WIN55212-2 was applied at the end of the experiment.

f, Summary of the peak change in eCB2.0 fluorescence in response to 20 pulses applied at baseline (Ctrl), 26 min after DO34 application, and after WIN55212-2 application; n = 3 cultures each, mean ± s.e.m. Paired two-tailed Student’s t test was performed: P=0.0004.

g, Diagram depicting the degradation pathways for 2-AG and AEA. AA, arachidonic acid; MAGL, monoacylglycerol lipase; FAAH, fatty acid amide hydrolase.

h, Representative traces (left) and expanded traces (right) showing the change in eCB2.0 fluorescence in response to 20 electrical pulses applied before (1) and after (2) JZL184 or URB597 application; AM251 was applied at the end of the experiment.

i, Summary of the decay time constant (τ_decay) measured at baseline (Ctrl) and 68 min after application of either JZL184 or URB597; n = 3 and 4 cultures each, mean ± s.e.m. Paired two-tailed Student’s t test was performed: P=0.0462 (left) and 0.0354 (right).

j, Pseudocolor images showing spontaneous changes in eCB2.0 fluorescence transients, single pulse–evoked fluorescence change, and the change in fluorescence induced by 10 μM WIN55212-2 (note the difference in scale). Scale bar, 100 μm.

k, Time-lapse pseudocolor images taken from the area shown by the bottom dashed rectangle in panel j. Scale bar, 10 μm.

l, Traces from the experiment shown in panel k, showing the change in fluorescence measured spontaneously, induced by a single pulse, or in the presence of AM251. Normalized traces with the corresponding rise time constants are shown at the right.

m, Spatial profile of the transient change in fluorescence shown in panel k. The summary data are shown at the right; n = 42 transients from 3 cultures.

n, Cumulative transient change in eCB2.0 fluorescence measured during 19 mins of recording in the absence (left) or presence (right) of AM251 (right). Pseudocolor images were calculated as the average temporal projection subtracted from the maximum temporal projection. Scale bar, 100 μm.

o, Summary of the frequency of transient changes in eCB2.0 fluorescence measured before (Ctrl) and after AM251 application; n = 5 & 3 with 10-min recording/session. Boxes show the first and third quartiles as well as the median (line), and the whiskers extend to the most extreme data point that is no more than 1.5× the interquartile range from the box. Two-tailed Student’s t test was performed: P=5.79E-06.

***, p < 0.001; *, p < 0.05.

Next, we asked which specific eCB—2-AG and/or AEA—is released in cultured rat cortical neurons. 2-AG is mainly produced in neurons from diacylglycerol (DAG) by diacylglycerol lipase (DAGL), while AEA is mainly produced from N-arachidonoyl phosphatidylethanolamine (NAPE) via the enzyme NAPE-hydrolyzing phospholipase D (NAPE-PLD) (Fig. 3d). We found that the selective DAGL inhibitor DO34⁴⁸ eliminated the stimulus-evoked eCB2.0 signal within 30 min; as a positive control, subsequent application of the CB1R agonist WIN55212-2 restored eCB2.0 fluorescence, indicating that the sensor is still present in the cell membrane (Fig. 3e,f). We also examined the effect of blocking the degradation of 2-AG and AEA via the enzymes monoacylglycerol lipase (MAGL) and fatty acid amide hydrolase (FAAH) using the inhibitors JZL184⁴⁹ and URB597⁵⁰, respectively (Fig. 3g). We found that blocking MAGL significantly increased the decay time constant, while blocking FAAH had only a slight—albeit significant—effect on the decay time constant (Fig. 3h,i). Taken together, these data indicate that 2-AG is the principal eCB released from cultured rat cortical neurons in response to electrical stimuli. Lastly, to estimate the concentration of 2-AG released from neurons, we recorded the fluorescence responses of eCB2.0 evoked by 1 to 100 electrical pulses at 20 Hz and then calibrated the sensor responses versus exogenous applied 2-AG with different concentrations. Using the calibration curve, we estimated that 0.5 – 7.7 μM 2-AG were released from cultured cortical neurons after electrical stimulation (Extended Data Fig. 5).

In addition to the stimuli-evoked eCB signals, we also observed local, transient eCB2.0 signals in neurons that occurred spontaneously in the absence of external stimulation (Fig. 3j). The peak amplitude and rise kinetics of these transient eCB2.0 signals were smaller and slower compared to the signal measured in response to a single electrical stimulus recording in the same region of interest (ROI) (Fig. 3k,l), suggesting that evoked and spontaneous eCB release have distinct patterns. The average diameter of the spontaneous transient signals was ~11 μm based on our analysis of full width at half maximum (FWHM) (Fig. 3m), consistent with previous suggestions that eCB acts locally^51,52, although the value might be a lower estimate limited by the sensor’s sensitivity. Finally, the spontaneous transient was no longer observed after the application of CB1R inverse agonist AM251 or CB1R neutral antagonist NESS0327, while the action potential blocker tetrodotoxin (TTX) didn’t block the spontaneous signals (Fig. 3l,n,o and Extended Data Fig. 6).

Imaging eCB in acute mouse brain slices

Next, we examined whether our eCB sensor can be used to detect endogenous eCB release in a more physiologically relevant system, namely acute mouse brain slices. We first injected AAVs expressing either eCB2.0 or eCBmut into the dorsolateral striatum (DLS) of adult mice (Fig. 4a), the region where eCB mediates both short-term and long-term depression and regulates motor behavior^53,54. Four weeks after AAV injection, acute brain slices were prepared, showing the expression of eCB sensors in DLS (Fig. 4b). The fluorescence signals evoked by electrical stimuli in the DLS were recorded by photometry. We found that applying electrical stimuli in eCB2.0-expressing slices evoked clear fluorescence signals, with stronger responses evoked by increasing the number of stimuli and by increasing the stimulation frequency (Fig. 4c,d). The half-rise time and decay time constant ranged from 0.8–1.2 s and 5.2–8.5 s (Fig. 4d). Moreover, the signal was specific to eCB release, as pretreating the slices with 10 μM AM251 blocked the response, and no response was measured in slices expressing the eCBmut mutant sensor (Fig. 4e). Under the 5-pulse 100-Hz stimulation condition, the eCB signal was mainly from 2-AG while AEA likely contributed to a small portion (Extended Data Fig. 7a-c). In a separate experiment, we performed electrophysiological recording of spontaneous inhibitory postsynaptic currents (sIPSCs) on medium spiny neurons (MSNs) in acute striatal slices, and observed eCB-mediated depolarization induced suppression of inhibition (DSI) (Extended Data Fig. 7d-f). Importantly, DSI was also successfully induced in eCB2.0 expressing striatal slices, suggesting eCB2.0 expression has minimal influence on the endogenous eCB-mediated modulation of synaptic plasticity (Extended Data Fig. 7d-f). We also expressed the eCB2.0 in the hippocampal CA1 and basal lateral amygdala (BLA) regions (Extended Data Fig. 8), where eCB modulates synaptic inputs^21,55,56, and then recorded eCB2.0 signals in acute slices using 2P microscopy. We found that applying an increasing number of electrical stimuli at 20 Hz evoked increasingly larger changes in eCB2.0 fluorescence (Extended Data Fig. 8b,c,g,h) in CA1 and BLA; 75 mM K⁺ perfusion induced a larger increase in eCB2.0 fluorescence in BLA (Extended Data Fig. 8i). In addition, applying 10 μM AEA to the hippocampal slices caused a profound increase in eCB2.0 fluorescence that was reversed by 10 μM AM251 (Extended Data Fig. 8d). Finally, AM251 eliminated the signal induced by even 100 field stimuli (Extended Data Fig. 8e).

Fig. 4 ∣. Using the GRAB_eCB sensor to detect eCB release in acute brain slices.

a, Schematic diagram depicting the strategy for virus injection in the dorsolateral striatum (DLS), followed by the preparation of acute brain slices used for electrical stimulation and photometry recording. The dashed box corresponds to the image shown in panel b.

b, Fluorescence image of a coronal slice prepared from a mouse following injection of AAV-syn-eCB2.0 in the DLS, with a diagram showing the electrode position and photometry recording. Similar results were observed for more than 6 slices. Scale bar, 1 mm.

c, Representative traces showing the change in eCB2.0 fluorescence evoked by 2, 5, or 10 electrical pulses applied at the indicated frequencies.

d, Peak change in eCB2.0 fluorescence (left), rise t_1/2 (middle), and decay time constant (right) plotted against stimulation frequency for 2, 5, and 10 pulses; n = 6 slices, mean ± s.e.m.

e, Representative traces (left) and summary of the peak change in eCB2.0 fluorescence (right) evoked by electrical pulses at the indicated frequency in slices expressing eCB2.0 in the absence or presence of AM251 and in slices expressing eCBmut; n = 4 slices for eCB2.0 and AM251 groups, n = 3 for eCBmut group, mean ± s.e.m.

f, Schematic diagram depicting the strategy for viral expression in the hippocampal CA1 of CB1R-P2A-FlpO mice and the expression of eCB2.0 in CB1R positive neurons.

g, Fluorescence images of eCB2.0 expressed in CA1 region in saline (left) and 75 mM K⁺ solution (middle) conditions. Pseudocolor images showing the change in eCB2.0 fluorescence after 75 mM K⁺ solution perfusion (right). Bottom images are taken from the area shown by the dashed yellow box in the upper panels. Similar results were observed for more than 5 slices. Scale bar, 20 μm (upper) and 5 μm (bottom).

h, Representative images and 3D illustration showing a bouton during baseline, spontaneous eCB release and 75 mM K⁺ induced eCB release conditions. Scale bar, 2 μm.

i, Spatial profile of the transient change in fluorescence from a single bouton shown in panel h. The summary data are shown at the right; n = 6 and 21 boutons, mean ± s.e.m.

j, (Left) Temporal profile of the transient change in fluorescence from a single bouton shown in panel h. (Right) The summary data of t_1/2 and peak ΔF/F₀; n = 6 and 21 boutons, mean ± s.e.m.

To further examine the spatial resolution that eCB2.0 imaging can reach, we injected the FlpO recombinase dependent AAV vectors into hippocampal CA1 of the in-house generated CB1R-P2A-FlpO mice, to sparsely expressed eCB2.0 in CB1R positive neurons. The acute slices were then prepared and imaged using 2P microscopy (Fig. 4f). Perfusing 75 mM K⁺ solution evoked a large fluorescence response of eCB2.0 in individua boutons (Fig. 4g), which are presynaptic terminal. In some cases, spontaneous eCB2.0 signals were recorded from single boutons (Fig. 4h). The diameter (FWHM) of signals from single boutons was about 1.2 μm (Fig. 4i), consistent with the size of a presynaptic terminus⁵⁷. The kinetics and peak fluorescent response were also quantified benefited from the temporal resolution and sensitivity of eCB2.0 (Fig. 4j). Together, these in vitro data confirm that eCB2.0 can be used to reliably detect the endogenous release of eCBs in acute brain slices with high sensitivity, specificity and spatiotemporal resolution.

eCB measurement in the BLA of freely moving mice

The basolateral amygdala (BLA) is a key brain region mediating fear responses and processing aversive memories⁵⁸. Previous studies found that the CB1R is highly expressed in the BLA, and the eCB system in BLA participates in stress expression^59-61. We therefore tested whether our eCB2.0 sensor can be used to directly measure eCB dynamics in vivo. First, we injected AAV vectors expressing eCB2.0 as well as the red-shifted channelrhodopsin ChRmine⁶² in the mouse BLA (Fig. 5a,b). Through the same optic fiber, we delivered red light to stimulation BLA neurons and simultaneously performed photometry recording of eCB2.0 at the same brain region in living mice. A 1-sec 40-Hz 635 nm light stimulation triggered a transient green fluorescence increase of eCB2.0 excited by 470 nm light but not 405 nm (isosbestic) light (Fig. 5c,d). Next, we injected AAV vectors expressing either eCB2.0 or eCBmut together mCherry in the mouse BLA and then performed fiber photometry recording while applying an aversive stimulus (electrical foot shock) (Fig. 5e-f). We found that applying a 2-sec foot shock induced a time-locked transient increase in eCB2.0 fluorescence in the BLA (Fig. 5g); this response was highly reproducible over 5 consecutive trials (Fig. 5h). Importantly, the same foot shock had no effect on either mCherry fluorescence or eCBmut fluorescence (Fig. 5i). The average time constant for the rise and decay phases of the eCB2.0 signal was 1.0 s and 6.3 s, respectively (Fig. 5j). To confirm the expression of eCB2.0 has no negative effect on animal behaviors, we performed the open field test, elevated plus maze test and fear conditioning, and found no behavioral differences between mice expressing GPF (as a control) or eCB2.0 in BLA bilaterally (Extended Data Fig. 9a-f). These data indicate that eCB2.0 can be used to measure eCB dynamics in vivo in freely moving animals without perturbing animal behaviors.

Fig. 5 ∣. Measuring in vivo eCB signals in the mouse basolateral amygdala in response to foot shock.

a, Schematic diagram depicting the strategy for viral expression in the basolateral amygdala, optogenetic stimulation and fiber photometry recording.

b, Immunofluorescence image showing eCB2.0 (green) and ChRmine (red) expressed in the BLA and the placement of the recording fiber. Similar results were observed for 4 mice. Scale bar, 200 μm.

c, Representative traces of the change in eCB2.0 fluorescence (averaged green trace with 5 single-trial gray traces) in response to 635 nm laser stimulation in the BLA from 1 mouse; isosbestic signals (averaged black trace with 5 single-trial gray traces) were used to monitor potential movement artifacts.

d, Quantification of peak amplitudes (z-score) of the signals in all mice. n = 4 mice, mean ± s.e.m. Paired two-tailed Student’s t test was performed: P=0.0418.

e, Schematic diagram depicting the strategy for viral expression in the basolateral amygdala.

f, Immunofluorescence image showing eCB2.0 (green) and mCherry (red) expressed in the BLA and the placement of the recording fiber; the nuclei were counterstained with DAPI (blue). Similar results were observed for 6 mice. Scale bar, 300 μm.

g, Schematic diagram depicting the fiber photometry recording during foot shock and representative single-trial traces of the change in eCB2.0 and mCherry fluorescence; an electrical foot shock (2-sec duration) was applied at time 0.

h, Pseudocolor change in eCB2.0 fluorescence before and after a 2-sec foot shock. Shown are five consecutive trials in one mouse, time-aligned to the onset of each foot shock.

i, (Left) average traces of the change in eCB2.0 and mCherry (top) and eCBmut and mCherry (bottom) fluorescence; the gray shaded area indicates application of an electrical foot shock. (Right) summary of the peak change in fluorescence; n = 6 mice each, mean ± s.e.m. Paired two-tailed Student’s t test (between eCB2.0 and mCh) and two-tailed Student’s t test (between eCB2.0 and eCBmut) were performed: P=0.0009 and 1.35E-06.

j, Summary of rise and decay time constants measured for the change in eCB2.0 fluorescence in response to foot shock; n = 21 trials for rise measurement and n = 18 trials for decay measurement from 6 animals, mean ± s.e.m.

***, p < 0.001; *, p < 0.05.

Imaging eCB in the mouse CA1 during running and seizures

Our finding that eCB2.0 can be expressed in the mouse hippocampal CA1 region and then measured in acute slices led us to ask whether we can use this sensor to measure in vivo eCB dynamics in the CA1 region during physiologically relevant activity such as running. We therefore injected AAVs expressing eCB2.0 or eCBmut together with a red Ca²⁺ indicator jRGECO1a⁶³ into mouse hippocampal CA1 region and then conducted head-fixed 2P dual-color imaging through an implanted cannula above the hippocampus (Fig. 6a). Co-expression of eCB2.0 and jRGECO1a was clearly observed in neurons in the CA1 4–6 weeks after virus injection (Fig. 6b). We focused on the stratum pyramidale layer, which is composed of pyramidal neuron somata and interneuron axons, including a class that densely express CB1R. When mice spontaneously ran on a treadmill (Fig. 6c), we found rapid increases of both calcium and eCB signals aligned to the start of running, and decreases of both signals when the running stopped (Fig. 6d,e). In the control group, which expressed eCBmut and jRGECO1a, calcium signals were intact while eCBmut showed no fluorescence change (Fig. 6d,e). Both calcium and eCB signals had similar 10%-90% rise time and half-time of the decay phase (Fig. 6f). Lastly, no difference was observed in running speed or distance among control, eCB2.0 expressing and eCBmut expressing mice (Extended Data Fig. 9g), further demonstrating eCB sensors have no detectable effect on animal behaviors.

Fig. 6 ∣. Measuring in vivo eCB dynamics in the mouse hippocampus during running and seizure activity.

a, Schematic diagram depicting the strategy for viral expression and cannula placement in the mouse hippocampus.

b, (Left) Fluorescence microscopy image showing eCB2.0 expression in the hippocampal CA1 region in a coronal brain slice. Scale bars, 200 μm and 50 μm (inset). (Right) In vivo 2-photon image of the pyramidal layer in the hippocampal CA1 region, showing eCB2.0 (green) and jRGECO1a (red) fluorescence. Scale bar, 50 μm.

c, Schematic cartoon illustrating the experiment in which a mouse expressing eCB2.0 and jRGECO1a in the hippocampal CA1 is placed on a treadmill and allowed to run spontaneously while fluorescence is measured using 2-photon microscopy.

d, Average traces of eCB2.0/eCBmut and jRGECO1a transients recorded in the soma of individual neurons in the pyramidal layer upon the start and stop of spontaneous running episodes (dashed lines).

e, Summary of the peak responses in panel d; n = 7 and 4 mice each for eCB2.0 and eCBmut, respectively, mean ± s.e.m. Two-tailed Student’s t test was performed between jR and jR: P=0.5979; one-tailed Student’s t test was performed between eCB2.0 and eCBmut: P=0.0284.

f, Summary of the rise and decay kinetics of the jRGECO1a and eCB2.0 signals measured at the start and end of spontaneous running; n = 7 mice, mean ± s.e.m. Paired two-tailed Student’s t tests were performed: P=0.5954 (left) and 0.0993 (right).

g, Schematic diagram depicting the electrode placement and 2-photon imaging in mice expressing eCB2.0 and jRGECO1a in the hippocampal CA1 region; the electrode is used to induce kindling seizure activity and to measure the local field potential (LFP).

h, Example LFP trace (top) and medio-lateral projections (line profile) of jRGECO1a (middle) and eCB2.0 (bottom) fluorescence during stimulus-induced non-convulsive seizures and a subsequent spreading wave. The dashed vertical line at time 0 indicates the stimulus onset.

i, Individual (thin gray lines) and average (thick lines) traces of the change in jRGECO1a and eCB2.0/eCBmut fluorescence measured during seizure activity. The dashed vertical line at time 0 indicates the stimulus onset. The summary of the area under the curve (AUC) is shown at the right; n = 8 and 4 for eCB2.0 and eCBmut, respectively, mean ± s.e.m. Two-tailed Student’s t tests were performed: P=0.2607 (between jR and jR:) and P=0.00098 (between eCB2.0 and eCBmut).

j, Spreading eCB wave measured through the hippocampal CA1 region after seizure activity. ROIs representing individual neurons are pseudocolored based on the peak time of their eCB2.0 signal relative to the peak time of the average signal, and the arrow shows the direction of the wave. a, anterior; l, lateral; m, medial; p, posterior.

k, Traces of eCB2.0 fluorescence measured in individual cells sampled systematically along a line fitted to the spreading wave. The dashed line shows the spreading of peak signals.

l, Velocity and direction of the spreading jRGECO1a and eCB2.0 waves. The length of each arrow indicates the velocity in μm/s. In each panel, each colored arrow indicates an individual session, and the thick black line indicates the average. n = 7 sessions in 6 mice.

***, p < 0.001; *, p < 0.05; n.s., not significant.

Epilepsy is a neurological disease characterized by excessive and synchronous neuronal firing. eCBs are proposed to provide negative feedback during epilepsy to attenuate the synaptic activity and protect the nervous system, which is exemplified by the observation that animals with compromised eCB system all exhibit a pro-epileptic phenotype⁶⁴. To explore whether our eCB2.0 sensor can be used to study seizure-related eCB signals in vivo, we used electrical kindling stimulation⁶⁵ of the hippocampus contralateral to the sensor expressing hemisphere to elicit brief self-terminating seizures (measured using local field potential (LFP) recording) (Fig. 6g). We found strong increases in both calcium and eCB signals during electrical seizure activity (Fig. 6h). Recent work has shown that seizures are often followed by a spreading calcium wave that propagates across the cell layer⁶⁵. Interestingly, we also found a propagating eCB wave that closely followed the calcium wave (Fig. 6h, Extended Data Fig.10 and Supplementary Video 1). In contrast, eCBmut showed no response during and after seizures (Fig. 6i). The velocity and direction of eCB waves were evident when we extracted the eCB2.0 signal from individual neurons in the field of view (Fig. 6j,k). Notably, eCB waves and calcium waves varied across experiment sessions and animals (Fig. 6l), but for each instance, the calcium and eCB waves were similar, in agreement with the calcium- and activity-dependence of the eCB signal. Taken together, our results confirm that the eCB2.0 sensor can be used to measure eCB dynamics in vivo under both physiological and pathological conditions, with high specificity and spatiotemporal resolution.

DISCUSSION

Here, we report the development and characterization of a genetically-encoded fluorescent sensor for detecting eCBs both in vitro and in vivo. With high sensitivity, selectivity and kinetics, this novel eCB sensor can be used to detect endogenous eCB release in cultured neurons, acute brain slices and in specific brain structures in vivo such as the amygdala and hippocampus during both physiological and pathological activities.

The dynamic range of eCB2.0 in neurons is higher than that in HEK293T cells, possibly because the membrane trafficking of eCB2.0 is better in neurons. Our estimate of τ_on and τ_off kinetics measured for eCB2.0 in cultured cells at room temperature is likely high, given that a faster time constant was measured in acute slices and in our in vivo experiments, the difference of which may come from the distinct experimental setup, instrumentation, temperature, and eCB source. Nevertheless, given that the temporal resolution of eCB2.0 is on the order of seconds, this tool is a vast improvement compared to microdialysis (with temporal resolution on the order of minutes), although the sensor’s kinetics can be improved even further in order to capture more rapid signals⁶⁶. eCB2.0’s apparent affinities for AEA and 2-AG are in submicromolar and micromolar range respectively. The relative lower affinity for 2-AG may prevent the sensor from detecting small amount of 2-AG. Improved sensors with higher affinities or larger dynamic range will be more useful in detecting mild eCB activities. In addition, the eCB2.0 sensor can detect both 2-AG and AEA; given that 2-AG and AEA regulate distinct pathways and are involved in different brain regions and cell types⁴, next-generation GRAB_eCB sensors should be developed with non-overlapping eCB specificity, as well as non-overlapping color spectra.

The retrograde modulation of synaptic activity by eCBs was previously identified by studying depolarization-induced suppression of inhibition (DSI) and excitation (DSE) in the hippocampus and cerebellum^21,24,26. However, because these experiments and subsequent studies used electrophysiological recordings of synaptic transmission combined with either pharmacological interventions or genetic manipulation, they lacked the ability to directly measure eCB release. Recording at the cell body of a neuron does not provide precise spatial information with respect to eCB release. For example, DSI recorded using paired whole-cell recordings in hippocampal slices indicates that depolarization of one neuron can inhibit GABAergic input to neurons within approximately 20 μm, suggesting the upper limit of diffusion for eCBs from a single neuron²¹; similar results were obtained in cerebellar slices using two separate stimulating electrodes to evoke eCB release from two dendritic regions in a single Purkinje cell⁵¹. Although these data indicate that eCB signaling is relatively localized and tightly controlled, the detailed spatial profile of eCB signaling is unknown. In addition, the eCB signals measured by changes in evoked postsynaptic currents have a sampling interval of approximately 2 s, creating a temporal bottleneck. In this respect, our eCB2.0 sensor can reveal eCB signals with considerably higher spatial and temporal resolution.

In summary, we show that our eCB2.0 sensor can be used in a variety of in vitro and in vivo preparations in order to monitor eCB dynamics in real time. Given the complexity of the nervous system, future directions for research based on the eCB sensor applications may include the identity of cell types that release eCBs, the mechanisms and temporal properties of eCB release, characteristics of eCB diffusion, the duration of eCB signals, the nature of the cell types and subcellular elements targeted by eCBs and the effects on them. Answering these fundamental questions will significantly enrich our understanding of the mechanisms and functions of eCB signaling at the synapse and neural circuit levels. Lastly, altered function of the eCB system has been associated with several neurological disorders, including stress/anxiety, movement disorders, substance use disorders and epilepsy. In this respect, our in vivo results show clear examples of how the eCB2.0 sensor can help to elucidate the fast eCB dynamics during both physiological and pathological processes. Thus, eCB sensors open a new era of endocannabinoid research aimed at understanding this system at unprecedented, physiologically-relevant spatial and temporal scales.

METHODS

Molecular biology

DNA fragments were amplified by PCR using primers (TSINGKE Biological Technology) with 25–30-bp overlaps. Plasmids were constructed using restriction enzyme cloning or Gibson Assembly, and all plasmid sequences were verified using Sanger sequencing. To characterize eCB2.0 and eCBmut in HEK293T cells, the corresponding DNA constructs were cloned into the pDisplay vector with an upstream IgK leader sequence. An IRES-mCherry-CAAX cassette was inserted downstream of the sensor gene for labeling the cell membrane and calibrating the sensor’s fluorescence. To characterize eCB2.0 in neurons, the eCB2.0 was cloned into a pAAV vector under control of a human synapsin (SYN1) promoter (pAAV-hSyn), and PSD95-mScarlet and synaptophysin-mScarlet were cloned into the pDest vector under the control of the CMV promoter. For the G_βγ sensor assay, the human CB1R was cloned into the pCI vector (Promega), and eCB2.0 and eCBmut were cloned into the peGFP-C1 vector (Takara), replacing the eGFP open reading frame. For the Tango assay, the human CB1R, eCB2.0 and eCBmut were cloned into the pTango vector. In addition, the viral vectors pAAV-hsyn-eCBmut and pAAV-hsyn-R^ncp-iGluSnFR were generated and used in this study.

AAV expression

AAV2/9-hSyn-eCB2.0 (9.5x10¹³ viral genomes (vg)/mL), AAV2/9-hSyn-eCBmut (8.0x10¹³ vg/mL), AAV2/9-EFS-fDIO-eCB2.0 (8.3 x10¹³ vg/mL), AAV2/9-hSyn-R^ncp-iGluSniFR (6.2x10¹³ vg/mL, all packaged at Vigene Biosciences, China), AAV9-CaMKII-ChRmine-mScarlet (K. Deisseroth’s lab at Stanford University), AAV8-hSyn-mCherry (#114472, Addgene), AAV9-CAG-GFP (University of North Carolina Vector Core Facility) and AAV1-Syn-NES-jRGECO1a-WPRE-SV40 (Penn Vector Core) were used to infect cultured neurons or were injected in vivo into specific brain regions.

Cell culture

HEK293T cells were cultured at 37°C in air containing 5% CO₂ in DMEM (Biological Industries) supplemented with 10% (v/v) fetal bovine serum (Gibco) and penicillin (100 unit/mL)-streptomycin (0.1 mg/mL) (Biological Industries). For experiments, the HEK293T cells were plated on 96-well plates or 12 mm glass coverslips in 24-well plates. At 60–70% confluency, the cells were transfected using polyethylenimine (PEI) with 300 ng DNA/well (for 96-well plates) or 1 μg DNA/well (for 24-well plates) at a DNA:PEI ratio of 1:3; 4–6 h after transfection, the culture medium was replaced with fresh medium. Imaging was performed 24–36 h after transfection. Rat cortical neurons were prepared from postnatal day 0 (P0) Sprague-Dawley rat. In brief, the cerebral cortex was dissected, and cortical neurons were dissociated by digestion in 0.25% Trypsin-EDTA (Biological Industries), and then plated on poly-D-lysine–coated (Sigma-Aldrich) 12-mm glass coverslips in 24-well plates. The neurons were cultured at 37°C, 5% CO₂ in Neurobasal Medium (Gibco) supplemented with 2% B-27 Supplement (Gibco), 1% GlutaMAX (Gibco), and penicillin (100 unit/mL)-streptomycin (0.1 mg/mL) (Biological Industries). For transfection experiment in Fig. 2a, cultured neurons were transfected at 7–9 day in vitro (DIV7–9) using calcium phosphate transfection method and imaged 48 h after transfection. For viral infection, cultured neurons were infected by AAVs expressing eCB2.0, eCBmut and/or R^ncp-iGluSnFR at DIV3–5 and imaged at DIV12–20. Where indicated, the neurons were loaded with Calbryte-590 (AAT Bioquest) 1 h before imaging.

Animals

All experiment protocols were approved by the respective Laboratory Animal Care and Use Committees of Peking University, the National Institute on Alcohol Abuse and Alcoholism, the Cold Spring Harbor Laboratory, and Stanford University, and all studies were performed in accordance with the guidelines established by the US National Institutes of Health. Postnatal day 0 (P0) Sprague-Dawley rats (Beijing Vital River Laboratory) of both sexes and P42–P150 C57BL/6J mice (Beijing Vital River Laboratory and The Jackson Laboratory) of both sexes were used in this study. CB1R-P2A-FlpO mice line was generated by inserting P2A-FlpO sequence before Cnr1 stop codon using CRISPR/Cas9 method and used in this study. The mice were housed at 18–23 °C with 40–60% humidity under a normal 12-h light/dark cycle with food and water available ad libitum.

Confocal imaging of cultured cells

Before imaging, the culture medium was replaced with Tyrode’s solution consisting of (in mM): 150 NaCl, 4 KCl, 2 MgCl₂, 2 CaCl₂, 10 HEPES, and 10 glucose (pH 7.4). 0 mM [Ca²⁺]_ex solution was modified from Tyrode’s solution with 0 mM CaCl₂ and additional 2 mM EGTA. HEK293T cells in 96-well plates were imaged using an Opera Phenix high-content screening system (PerkinElmer, USA) equipped with a 20x/0.4 NA objective, a 40x/0.6 NA objective, a 40x/1.15 NA water-immersion objective, a 488 nm laser and a 561 nm laser. Green and red fluorescence were collected using a 525/50 nm emission filter and a 600/30 nm emission filter, respectively. Cells in 12 mm coverslips were imaged using a Ti-E A1 confocal microscopy (Nikon, Japan) equipped with a 10x/0.45 NA objective, a 20x/0.75 NA objective, a 40x/1.35 NA oil-immersion objective, a 488 nm laser and a 561 nm laser. Green and red fluorescence were collected using a 525/50 nm emission filter and a 595/50 nm emission filter, respectively. The following compounds were applied by replacing the Tyrode’s solution (for imaging in 96-well plates) or by either bath application or using a custom-made perfusion system (for imaging cells on 12-mm coverslips): 2-AG (Tocris), AEA (Cayman), AM251 (Tocris), LPA (Tocris), S1P (Tocris), ACh (Solarbio), DA (Sigma-Aldrich), GABA (Tocris), Glu (Sigma-Aldrich), Gly (Sigma-Aldrich), NE (Tocris), 5-HT (Tocris), His (Tocris), Epi (Sigma-Aldrich), Ado (Tocris), Tyr (Sigma-Aldrich), WIN55212-2 (Cayman), CP55940 (Cayman), Δ-9-THC (The Third Research Institute of Ministry of Public Security), NESS0327 (Cayman), TTX (Must Biotech), DO34 (MedChemExpress), JZL184 (Cayman), and URB597 (Cayman). The micropressure application of drugs was controlled by Pneumatic PicoPump PV800 (World Precision Instruments). Cultured neurons were field stimulated using parallel platinum electrodes positioned 1 cm apart; the electrodes were controlled by a Grass S88 stimulator (Grass Instruments), and 1-ms pulses were applied at 80 V. All imaging experiments were performed at room temperature (22–24°C).

Fluorescent spectra and photostability measurement

Plasmid expressing eCB2.0 or EGFP-CAAX was transfected into HEK293T cells in 6-well plates (for 1P spectra measurement) or on 12-mm coverslips (for 2P spectra and photostability measurement). For 1P spectra measurement, approximately 24 – 36 hours after transfection, cells were digested by trypsin, washed by PBS, resuspended in Tyrode’s solution (or Tyrode’s solution containing 100 μM 2-AG), and placed into 384-well plates., in which each well contained cells from 1/4 of cells from a 6-well plate well. Control cells transfected with an empty vector were used for background subtraction. Using the Safire 2 plate reader (TECAN), excitation spectra were measured from 350 – 520 nm with a 5-nm step size and 20-nm bandwidth; the emission was set at 560 nm with a 20-nm bandwidth. Emission spectra were measured from 500 – 650 nm with a 5-nm step size and 20-nm bandwidth; the excitation was set at 455 nm with a 20-nm bandwidth. For 2P spectra measurement, cells in Tyrode’s solution or Tyrode’s solution containing 100 μM 2-AG were imaged using an Ultima Investigator 2P microscopy (Bruker) equipped with a 20x/1.00 NA objective (Olympus) and an InSight X3 tunable laser (Spectra-Physics). Images were captured when cell were excited from 700 – 1050 nm with a 10-nm step size. The fluorescence of untransfected cells was subtracted as background. The laser power was normalized according to the output power of the tunable 2P laser with different wavelength. Photostability was measured under 1P illumination (confocal laser scanning) using a 488 nm laser with laser power of ~100 μW; and under 2P illumination using a 920 nm laser with laser power of ~100 mW.

G_q-i Ca²⁺ imaging assay

Plasmids expressing eCB2.0 or CB1R-EGFP were co-transfected into HEK293T cells with a single construct expressing jRGECO1a-P2A-G_q-i1. Approximately 24 – 36 hours after transfection, cells in 12 mm coverslips were imaged using a Ti-E A1 confocal microscopy (Nikon, Japan) as described above. 2-AG with indicated concentrations was applied by bath application and washed using a custom-made perfusion system.

BRET G_βγ sensor assay

The BRET G_βγ sensor was developed based upon similar systems^67,68. Plasmids expressing eCB2.0, eCBmut, or CB1R were co-transfected into HEK293T cells together with a single construct expressing human GNAOa, human GNB1 (fused to amino acids 156–239 of Venus), human GNG2 (fused to amino acids 2–155 of Venus), and NanoLuc fused to the amino terminal 112 amino acids of human Phosducin circularly permutated at amino acids 54/55 (Promega). The NanoLuc/Phosducin fusion portion also contains a kRAS membrane targeting sequence at the carboxy terminal end. Templates for assembly were derived from human whole-brain cDNA (Takara) for all cDNAs, except for the hGNB1 and hGNG2 Venus fusions which were a generous gift from Dr. Nevin Lambert (Augusta University). Approximately 24 hours after transfection, the cells were harvested with 10 mM EDTA in phosphate-buffered saline (PBS, pH 7.2), pelleted, and then resuspended in Dulbecco’s modified PBS (Life Technologies) without Ca²⁺ or Mg²⁺. Furimazine (Promega) was then added at a 1/100 dilution to 100 μL of cell suspension in a black 96-well plate, and BRET was measured using a PHERAstar FS plate reader (Berthold) equipped with a Venus BRET cube. The acceptor (Venus) and donor (NanoLuc) signals were measured at 535 nm and 475 nm, respectively, and net BRET was calculated by subtracting the acceptor/donor ratio of a donor-only sample from the acceptor/donor ratio of each sample. Readings were taken before and 3–4 min after application of 20 μM 2-AG (Tocris) to activate CB1R or the eCB sensor.

Tango assay

Plasmids expressing eCB2.0, eCBmut, or CB1R were transfected into a reporter cell line expressing a β-arrestin2-TEV fusion gene and a tTA-dependent luciferase reporter gene⁴⁵. 24 h after transfection, cells in 6 well plates were collected after trypsin digestion and plated in 96 well plates. AEA was applied at final concentrations ranging from 0.01 nM to 10 μM. 12 h after luciferase expression, Bright-Glo (Fluc Luciferase Assay System, Promega) was added to a final concentration of 5 μM, and luminescence was measured using the VICTOR X5 multi-label plate reader (PerkinElmer).

Photometry recording in the dorsolateral striatum in acute mouse brain slices

Adult (>10 weeks of age) male C57BL/6J mice were anesthetized with isoflurane, AAV vectors were injected (300 nl at a rate of 50 nl/min) into the dorsolateral striatum at the following coordinates relative to Bregma: A/P: +0.75 mm; M/L: ±2.5 mm; and D/V: −3.5 mm). After virus injection, the mice received an injection of ketoprofen (5 mg/kg, s.c.), and postoperative care was provided daily until the mice regained their preoperative weight. After a minimum of 4 weeks following AAV injection, the mice were deeply anesthetized with isoflurane, decapitated, and the brains were removed and placed in ice-cold cutting solution containing (in mM): 194 sucrose, 30 NaCl, 4.5 KCl, 26 NaHCO₃, 1.2 NaH₂PO₄, 10 D-glucose, and 1 MgCl₂ saturated with 5% CO₂/95% O₂. Coronal brain slices (250-μm thickness) were prepared and then incubated at 32°C for 60 min in artificial cerebrospinal fluid (ACSF) containing (in mM): 124 NaCl, 4.5 KCl, 26 NaHCO₃, 1.2 NaH₂PO₄, 10 D-glucose, 1 MgCl₂, and 2 CaCl₂. After incubation at 32°C, the slices were kept at room temperature until use. Photometry recordings were acquired using an Olympus BX41 upright epifluorescence microscope equipped with a 40x/0.8 NA water-emersion objective and a FITC filter set. Slices were superfused at 2 mL/min with ACSF (29–31°C). A twisted bipolar polyimide-coated stainless-steel stimulating electrode (~200 μm tip separation) was placed in the DLS just medial to the corpus callosum and slightly below the tissue surface in a region with visible eCB2.0 or eCBmut fluorescence. The sensors were excited using either a 470-nm light-emitting diode (LED) (ThorLabs). Photons passing through a 180-μm² aperture positioned just lateral to the stimulating electrode were directed to a model D-104 photomultiplier tube (PMT) (Photon Technology International). The PMT output was amplified (gain: 0.1 μA/V; time constant: 5 ms), filtered at 50 Hz, and digitized at 250 Hz using a Digidata 1550B and Clampex software (Molecular Devices). For each photometry experiment, GRAB_eCB was measured as discrete trials repeated every 3 minutes. For each trial, the light exposure duration was 35–45 seconds in order to minimize GRAB_eCB photobleaching while capturing the peak response and the majority of the decay phase. To evoke an eCB transient, a train of 200–500-μs electrical pulses (1.0–1.5 mA) was delivered 5 s after initiating GRAB_eCB excitation.

Electrophysiology recordings of DSI in acute striatal slices

Adult (5–8 weeks of age) C57BL/6J mice of both sexes were anesthetized with an intraperitoneal injection of 2,2,2-tribromoethanol (Avertin, 500 mg/kg body weight, Sigma-Aldrich), and AAV vectors were injected (300 nl at a rate of 50 nl/min) into the dorsal striatum region using the following coordinates (relative to bregma): A/P: 1.0 mm; M/L: 2.35 mm; and D/V: −3.4 mm. After 4 weeks following AAV injection, acute brain slices were prepared for recording. Mice were anesthetized with isoflurane, decapitated, and brains were extracted and briefly submerged into chilled artificial cerebrospinal fluid (ACSF) containing (in mM): 125 NaCl, 2.5 KCl, 1.25 NaH₂PO₄, 25 NaHCO₃, 15 glucose, 2 CaCl₂, and 1 MgCl₂, oxygenated with 95% O₂ and 5% CO₂ (300-305 mOsm, pH 7.4). Oblique horizontal slices (300 μm thickness) containing dorsal striatum were then prepared using a tissue vibratome (VT1200S, Leica), incubated in chambers containing 34°C ACSF for 30 min, and then allowed to recover at room temperature for 30 min. After recovery, slices were transferred to a submerged recording chamber perfused with ACSF at a rate of 2-3 mL/min at a temperature of 30-31°C. All recordings experiments were performed within 5 hrs of slice recovery. Whole-cell voltage clamp DSI recordings were made with glass pipettes (3-4 MΩ) filled with a high-chloride internal solution including (in mM): 125.2 CsCl, 10 NaCl, 10 HEPES, 1 EGTA, 2 QX-314 chloride, 0.1 CaCl₂, 4 Mg-ATP, 0.3 Na₃-GTP, and 8 disodium phosphocreatine (280-290 mOsm, pH 7.3 with CsOH). For DSI recordings, a high-Ca²⁺ ACSF (4 mM Ca²⁺, 0.5 mM Mg²⁺) as perfused to increase the rate of spontaneous synaptic activity. NBQX (10 μM) and R-CPP (10 μM) were included in the perfusion to block AMPAR- and NMDAR-mediated currents respectively. Cells were voltage clamped at −70 mV. Access resistance was measured by injection of hyperpolarizing pulses (−5 mV, 100 μs) and was less than 25 MΩ for all recordings and only cells with a change in access resistance <20% throughout the entire experiment were included in the analysis. sIPSCs were recorded for a baseline of 10 seconds before depolarization to 0 mV for 5 seconds and additional recording of sIPSCs for 20 seconds after depolarization (Wilson and Nicoll, 2001). sIPSC charge (integrated current) was binned every 2 seconds, normalized to the average of the 10 seconds (5 bins) preceding depolarization, and the normalized charge before depolarization, after depolarization, and 16 seconds after depolarization were compared. Whole-cell patch clamp recordings were performed using a Multiclamp 700B (Molecular Devices), monitored with WinWCP (Strathclyde Electrophysiology Software) and analyzed offline using Clampfit 10.0 (Molecular Devices) and custom-made MATLAB (Mathworks) software. Signals were filtered at 2 kHz and digitized at 10 kHz (NI PCIe-6259, National Instruments).

2-photon imaging in the hippocampus and BLA in acute mouse brain slices

Adult (6–8 weeks of age) C57BL/6J mice or adult CB1R-P2A-FlpO mice (6–8 weeks of age) of both sexes were anesthetized with an intraperitoneal injection of 2,2,2-tribromoethanol (Avertin, 500 mg/kg body weight, Sigma-Aldrich), and AAV vectors were injected (400 nl at a rate of 46 nl/min) into the hippocampal CA1 region (coordinates: A/P: −1.8 mm relative to Bregma; M/L: ±1.0 mm relative to Bregma; and D/V: −1.2 mm relative to brain surface) or BLA region (coordinates: A/P: −1.40 mm relative to Bregma; M/L −3.10 mm relative to Bregma; and D/V: −4.20 mm relative to brain surface). After at least 4 weeks following AAV injection, the mice were deeply anesthetized with an intraperitoneal injection of 2,2,2-tribromoethanol, decapitated, and the brains were removed and placed in ice-cold cutting solution containing (in mM): 110 choline-Cl, 2.5 KCl, 0.5 CaCl₂, 7 MgCl₂, 1 NaH₂PO₄, 1.3 Na ascorbate, 0.6 Na pyruvate, 25 NaHCO₃, and 25 glucose saturated with 5% CO₂/95% O₂. Coronal brain slices (300-μm thickness) were prepared and incubated at 34°C for approximately 40 min in modified ACSF containing (in mM): 125 NaCl, 2.5 KCl, 2 CaCl₂, 1.3 MgCl₂, 1 NaH₂PO₄, 1.3 Na ascorbate, 0.6 Na pyruvate, 25 NaHCO₃, and 25 glucose saturated with 5% CO₂/95% O₂; 75 mM K⁺ ACSF instead contained 52.5 NaCl and 75 KCl. Two-photon imaging were performed using an FV1000MPE 2-photon microscope (Olympus) equipped with a 25x/1.05 NA water-immersion objective and a mode–locked Mai Tai Ti:Sapphire laser (Spectra-Physics). The slices were superfused with modified ACSF (32–34°C) at a rate of 4 mL/min. A 920-nm laser was used to excite the eCB2.0 sensor, and fluorescence was collected using a 495–540-nm filter. For electrical stimulation, a bipolar electrode (cat. number WE30031.0A3, MicroProbes for Life Science) was positioned near the stratum radiatum layer in the CA1 region using fluorescence guidance. Fluorescence imaging and electrical stimulation were synchronized using an Arduino board with custom-written software. The stimulation voltage was 4–6 V, and the pulse duration was 1 ms. Drugs were applied to the imaging chamber by perfusion at a flow rate at 4 mL/min.

Fiber photometry recording of eCB signals in the basolateral amygdala

Adult (10–12 weeks of age) C57BL/6J mice of both sexes were anesthetized, and 300 nl of either a 10:1 mixture of AAV-hSyn-eCB2.0 and AAV-hSyn-mCherry, a 10:1 mixture of AAV-hSyn-eCBmut and AAV-hSyn-mCherry, or a 1:1 mixture of AAV-hSyn-eCB2.0 and AAV-CaMKII-ChRmine-mScarlet was injected using a glass pipette and a Picospritzer III microinjection system (Parker Hannifin) into the right basolateral amygdala using the following coordinates: A/P: −1.78 mm relative to Bregma; M/L −3.30 mm relative to Bregma; and D/V: −4.53 mm relative to brain surface. After injection, a 200-μm diameter, 0.37 NA fiber (Inper) was implanted at the same location and secured using resin cement (3M). A head bar was also mounted to the skull using resin cement.

At least 14 days after surgery, photometry recording was performed using a commercial photometry system, FP3001 or FP3002 (Neurophotometrics). A patch cord (0.37 NA, Doric Lenses) was attached to the photometry system and to the fiber secured in the mouse brain. A 470-nm LED (160 μW) was used to excite the GRAB_eCB sensors, a 405 nm LED (50 μW) was used for excitation to obtain the isosbestic signal; and a 560-nm LED (25 μW) was used to excite mCherry.

For optogenetic manipulation experiments, mice were tested under head-restraint in a sound attenuated box. A 635 nm laser (1–3 mW) on the FP3002 system was used for optogenetic stimulation; a 40 Hz (5 ms per pulse) 1 s laser stimulation were delivered though the same fiber as that for photometry recording with an interval of 90 – 120 s between trials. For the foot shock experiments, the mice were allowed to move freely in a Habitest shock box (Coulbourn Instruments) inside a lighted soundproof behavior box. The FreezeFrame software program was used to apply triggers to the shock generator (Coulbourn Instruments). Five 2-sec pulses of electricity at an intensity of 0.7 mA were delivered to the shock box, with an interval of 90–120 s between trials. Photometry data were acquired with Bonsai 2.3.1 or 2.6.2 software (Bonsai), and were exported to MATLAB for further analysis. Photometry signals and behavioral events were aligned based on an analog TTL signal generated by the Bpod. To correct for photobleaching of fluorescence signals (baseline drift), a bi-exponential curve was fit to the raw fluorescence trace and subtracted as follows:

$$F_{raw\_fit}=fit\left(Timestamp,F_{raw},’\text{exp}2’\right)$$

$$F_{raw\_correction}=\frac{F_{raw}-F_{raw\_fit}}{F_{raw\_fit}}$$

After baseline drift correction, the fluorescence signals were z-scored relative to the mean and standard deviation of the signals in a time window −5 to 0 s relative to stimulus onset.

After photometry recording, the animals were deeply anesthetized and perfused with PBS followed by 4% paraformaldehyde (PFA) in PBS. The brains were removed, fixed in 4% PFA overnight, and then dehydrated with 30% sucrose in PBS for 24 h. Brain slices were cut using a Leica SM2010R microtome (Leica Biosystems). Floating brain slices were blocked at room temperature for 2 h with a blocking solution containing 5% (w/v) BSA and 0.1% Triton X-100 in PBS, and then incubated at 4°C for 24 h in PBS containing 3% BSA, 0.1% Triton X-100, and the following primary antibodies: chicken anti-GFP (1:1000, Aves, #GFP-1020) and rabbit anti-RFP (1:500, Rockland, #600-401-379). The next day, the slices were rinsed 3 times in PBS and incubated in PBS with DAPI (5 μg/mL, Invitrogen, #D1306) and the following secondary antibodies at 4°C for 24 h: Alexa Fluor 488 donkey anti-chicken (1:250, Jackson ImmunoResearch, #703-545-155) and Alexa Fluor 568 donkey anti-rabbit (1:250, Invitrogen, #A10042). Confocal images were captured using an LSM780 confocal microscope (Zeiss).

Behavioral tests in mice expressing eCB2.0 in BLA

Adult (10–12 weeks of age) C57BL/6J mice of both sexes were anesthetized, and 300 nl of either AAV-hSyn-eCB2.0 or AAV-CAG-GFP was injected into the BLA bilaterally similar to the BLA fiber photometry recording experiment. The same cohort of mice were subjected to various behavioral tests at least 14 days after surgery, with the order of tests being open field test, elevated plus maze test and fear conditioning. Mice were handled for two days, 10 min for each day before the experiments.

Open field test

The open field test was performed in a non-transparent square box (42.5 cm x 42.5 cm x 40 cm), which was enclosed in a sound-attenuating chamber illuminated with a house light. Animals were placed in one corner of the arena at the start of a test. Mice were allowed to explore the arena for 5 minutes while their behavior was videotaped at 5 Hz using a Logitech C930e camera. The arena was thoroughly cleaned with 70% ethanol in between subjects. The data were analyzed using Ethovision XT 5.1 software (Noldus Information Technologies, RRID:SCR_000441). The center zone was set to 21 × 21 cm in the middle of the arena.

Elevated plus maze test

The elevated plus maze has two open arms without walls (30 cm long and 5 cm wide) and two arms enclosed by 15.25-cm-high non-transparent walls. The arms were connected by a central platform (5 × 5 cm), with the identical (open or closed) arms being opposite to each other. The maze was 50 cm above the floor. At the start of the session, animals were first placed in the center zone, with their heads oriented to a closed arm. Mice were allowed to explore the maze for 10 minutes while their behavior was videotaped at 30 Hz using a Logitech C920 camera. The maze was thoroughly cleaned with 70% ethanol in between subjects. The data were analyzed using Ethovision XT 5.1 software.

Auditory fear conditioning

We followed standard procedures for conventional auditory fear conditioning ^69-71. Briefly, on day 1 (habituation), mice were handled and habituated to a conditioning cage, which was a Mouse Test Cage (18 cm x 18 cm x 30 cm) with an electrifiable floor connected to a H13-15 shock generator (Coulbourn Instruments, Whitehall, PA). The Test Cage was placed inside a sound-attenuating cabinet (H10-24A; Coulbourn Instruments) and illuminated with white light. During habituation, five repetitions of a sound (conditioned stimulus (CS); 7-kHz pure tone, 70 dB, 30 s in duration) were presented at 60–90s inter-trial intervals (ITIs).

On day 2 (conditioning), the mice were put back into the Test Cage inside the white-light-illuminated cabinet, as on day 1, and received five presentations of the CS, each of which co-terminated with a foot shock (unconditioned stimulus (US); 2 s in duration, 0.7 mA). The ITIs were 60-90s. Before each habituation and conditioning session, the Test Cage was wiped with 70% ethanol.

On day 3, the test for fear memory (retrieval) was performed in a novel context, where mice were exposed to two presentations of the CS, with an interval of 115 s. The novel context was a cage with a different shape (22 cm x 22 cm x 21 cm) and floor texture compared with the conditioning cage, and was illuminated with infrared light. Prior to each use, the floor and walls of the cage were wiped clean with 0.5% acetic acid to make the scent distinct from that of the conditioning cage.

Animal behavior was videotaped at 3.7 Hz with a monochrome CCD-camera (Panasonic WV-BP334). The FreezeFrame software (Coulbourn Instruments) was used to control the delivery of both tones and foot shocks. Freezing behavior was analyzed with FreezeFrame software.

2-photon in vivo imaging

Adult (100–150 days of age) C57BL/6J mice of both sexes were used for these experiments. The mice were anesthetized, and a mixture of AAV1-Syn-NES-jRGECO1a-WPRE-SV40 and either AAV9-hSyn-eCB2.0 or AAV9-hSyn-eCBmut (300–400 nl each, full titer) was injected into the right hippocampal CA1 region at the following coordinates relative to Bregma using a Hamilton syringe: A/P: 2.3 mm; M/L: 1.5 mm; and D/V: ‒1.35 mm. After virus injection, a stainless-steel cannula with an attached coverglass was implanted over the hippocampus as described previously^72,73, and a stainless-steel head bar was attached. A chronic bipolar wire electrode (tungsten, 0.002”, 0.5-mm tip separation, A-M Systems) was implanted into the left ventral hippocampus at the following coordinates relative to Bregma as previously described⁷⁴: A/P: 3.2 mm; M/L: 2.7 mm; and D/V: ‒4.0 mm. Head-fixed mice running on a linear treadmill with a 2-m-long cue-less belt were imaged using a resonant scanning 2-photon microscope (Neurolabware) equipped with a pulsed IR laser tunned to 1000 nm (Mai Tai, Spectra-Physics), GaAsP PMT detectors (H11706P-40, Hamamatsu), and a 16x/0.8 NA water-immersion objective (Nikon). The 2-photon image acquisition and treadmill speed were controlled and monitored using a Scanbox (Neurolabware). Bipolar electrodes were recorded using a model 1700 differential amplifier (A-M Systems). Seizures were elicited by applying an electric stimulation above the seizure threshold by 150 μA of current delivered in 1-ms biphasic pulses at 60 Hz for 1 s, using a model 2100 constant-current stimulator (A-M Systems). Following the in vivo recordings, the mice were anesthetized with isoflurane followed by an intraperitoneal injection of a mixture of ketamine (100 mg/kg body weight) and xylazine (10 mg/kg body weight) in saline. The mice were transcardially perfused with 0.9% NaCl for 1 min followed by 4% PFA and 0.2% picric acid in 0.1 M phosphate buffer. The brains were removed, post-fixed in the same fixative solution for 24 h at 4°C, then sliced on a VTS1200 vibratome (Leica Biosystems). The sections were then washed and mounted using VECTASHIELD (Vector Laboratories). Confocal images were acquired using an LSM710 imaging system equipped with a 20x/0.8 NA objective (Zeiss).

Data processing

Confocal imaging

Data for 96-well plate imaging were collected and analyzed using Harmony high-content imaging and analysis software (PerkinElmer). In brief, membrane regions were selected as regions of interest (ROIs) and the green fluorescence channel (i.e., the sensor) was normalized to the red fluorescence channel corresponding to mCherry-CAAX (G/R). ΔF/F₀ was then calculated using the formula [(G/R_drug – G/R_baseline)/(G/R_baseline)]. For 12-mm coverslip imaging, data were collected using the NIS-Element software (Nikon) and analyzed using ImageJ software (National Institutes of Health). ΔF/F₀ was calculated as using the formula [(F_t − F₀)/F₀], with F₀ representing baseline fluorescence. Data were plotted using OriginPro 2020 (OriginLab).

Slice photometry and 2-photon imaging

For slice photometry, GRAB_eCB signals were calculated as ΔF/F₀ by averaging the PMT voltage (V) for a period of 1 s just prior to electrical stimulation (F₀) and then calculating [V/(F₀-1)] for each digitized data sample. The decay phase was fitted with a single exponential, accounting for a sloping baseline. Rise t_1/2 was calculated in Prism v. 8.3(GraphPad) by fitting the rising phase of the signal with an asymmetrical logistics curve. Photometry sweeps were exported to Microsoft Excel 2016 to calculate normalized ΔF/F₀ traces and peak ΔF/F₀ values. For 2-photon imaging of slices, data were collected using FV10-ASW software (Olympus) and analyzed using ImageJ. ΔF/F₀ was calculated using the formula [(F_t – F₀)/F₀], with F₀ representing baseline fluorescence. Data were plotted using OriginPro 2020.

Fiber photometry in mice during foot shock

The fiber photometry data were analyzed off-line using Excel or MATLAB software (MathWorks) and plotted using OriginPro 2020.

2-photon imaging in mice during locomotion and seizure

Imaging data were processed and analyzed using Python scripts. To analyze single-cell responses, movies were initially motion-corrected using rigid translation, followed by non-rigid correction (HiddenMarkov2D) using the sima package⁷⁵. Binary ROIs were selected using a semi-automated approach. For the initial automated detection, movies were divided into segments consisting of 100 frames each; the average intensity projection of each segment was then computed, and the resulting resampled movie was used for detection. In sessions with electric stimulation, only the baseline period (i.e., before stimulation) was used for segmentation. The PlaneCA1PC method of sima was run on the inverted resampled movie, which resulted in detection of the hollow cell nuclei. These ROIs were then filtered based on size, and binary dilation was performed to include the cytoplasm around the nuclei. Next, the ROIs were detected in the non-inverted resampled movie and filtered based on size; those samples that did not overlap with existing ROIs were added to the set. ROIs outside the stratum pyramidale layer were excluded. The fluorescence intensity traces were then extracted for each ROI by averaging the included pixel intensities within each frame. For analyzing the run responses, only sessions with no electric stimuli were included, and signals were pulled from the motion-corrected movies. These raw traces were then processed following standard steps for obtaining ΔF/F₀ traces, with a modified approach for determining the time-dependent baseline. A 3rd-degree polynomial was fit to the trace after applying temporal smoothing, removing peaks (detected using continuous wavelet transform with scipy.signal), eliminating periods of running, and ignoring the beginning and end of the recording. The calculated polynomial was then used as a baseline. Z-scored traces were obtained after determining the standard deviation (SD) of each cell’s baseline and excluding events exceeding 2 SDs in two iterations.

To analyze spreading activity, only sessions with an electric stimulus that triggered an electrographic seizure and a spreading wave were included. The segmentation was performed based on the motion-corrected baseline segments of the recordings, and the signals were pulled from non-motion-corrected movies, as image-based motion correction was not feasible during seizures. ΔF/F₀ traces were obtained using a constant baseline determined by averaging the pre-stimulus segments of the traces. To analyze changes in average fluorescence intensity, a single large ROI was manually drawn to include the cell bodies within the pyramidal layer, and ΔF/F₀ traces were obtained and processed as described above. Event-triggered averages were calculated after automatically detecting the frames with running onsets and stops using criteria that were fixed across all sessions. The average was computed in two steps; first, the events were averaged by cell, and then the cells were averaged by sensor (e.g., eCB2.0 or eCBmut). Decay time constants were computed as the parameter of a 2nd-degree polynomial fit after a log transform on the trace following the peak of the stop-triggered average trace. Rise times were determined between the frame in which the start-triggered average signal first reached 90% of the range between baseline and peak and the last frame before the signal dropped below 10% of the range. To determine the speed and direction of the spreading waves, the peak time of the wave was determined in each session by inspecting the average ΔF/F₀ trace (including all cells). Next, the relative peak location (Δt) of the ΔF/F₀ trace of each cell in the trace including 200 frames (12.8 s) before and after the wave peak was determined. Finally, two linear (i.e., 1D) fits were determined using the x and y centroid coordinates of each ROI (Δt ~ x, Δt ~ y). The 2D speed was then computed from the slopes of the two 1D fits. The direction was determined by computing the unity vector from the starting point to the end point of the fits between 3 s before and after the wave peak. The average speed was obtained by averaging the speed of individual sessions, and the average direction was obtained from the sum of the unity vectors of individual sessions. Data were plotted using Python and OriginPro 2020.

Statistical analysis

Group data were analyzed using the Student’s t test, one-way ANOVA, two-way ANOVA, Mann-Whitney test or Wilcoxon matched-pairs signed rant tests. *, P<0.05, **, P<0.01, ***, P<0.001, and n.s., not significant (P>0.05). The exact P value is specified in the legends.

Extended Data

Extended Data Fig. 1. Strategy for optimizing and screening the GRAB_eCB sensors.

a, A flowchart showing the development process of the eCB2.0 sensor. Responses to 10 μM 2-AG of candidate sensors were shown alongside each step.

b, Schematic diagram depicting the structure of the GRAB_eCB2.0 sensor. The IgK leader sequence and the sequence derived from GRAB_NE are shown.

c, Amino acids sequence of the eCB2.0 sensor. The phenylalanine residue at position 177^2.64 in the CB1R was mutated to an alanine to generate the eCBmut sensor (indicated by the gray box). Note that the numbering used in the figure corresponds to the start of the IgK leader sequence.

Extended Data Fig. 2. Dose–response curves of GRAB_eCB2.0 to synthetic CB1R agonists and the phytocannabinoid Δ-9-THC.

a, Dose–response curve of eCB2.0 to WIN55212-2; n = 3 wells each, mean ± s.e.m.

b, Dose–response curve of eCB2.0 to CP55940; n = 3 wells each, mean ± s.e.m.

c, Dose–response curve of eCB2.0 to Δ-9-THC; n = 3 wells each, mean ± s.e.m.

Extended Data Fig. 3. Photostability and intracellular signaling couplings of GRAB_eCB2.0 sensor.

a, Normalized fluorescence of EGFP-CAAX and eCB2.0 (in the absence and presence of 2-AG) in HEK293T cells during 1P (confocal) bleaching.

b, Integrated fluorescence of EGFP-CAAX and eCB2.0 (in the absence and presence of 2-AG) shown in a; n = 29, 27, 28 cells from 3 cultures. Boxes show the first and third quartiles as well as the median (line), and the whiskers extend to the most extreme data point that is no more than 1.5× the interquartile range from the box. Two-tailed Mann-Whitney tests were performed: P=1.44E-10 (between EGFP and eCB2.0 in saline) and 1.37E-6 (between EGFP and eCB2.0 with 2-AG).

c, Fast and slow time constants and slow component amplitudes of EGFP-CAAX and eCB2.0 (in the absence and presence of 2-AG) traces fit by double exponentials.

d, Normalized fluorescence of EGFP-CAAX and eCB2.0 (in the absence and presence of 2-AG) in HEK293T cells during 2P bleaching.

e, Time constants of EGFP-CAAX and eCB2.0 (in the absence and presence of 2-AG) traces fit by single exponentials; n = 79, 48, 104 cells from 3 cultures. Boxes show the first and third quartiles as well as the median (line), and the whiskers extend to the most extreme data point that is no more than 1.5× the interquartile range from the box. Two-tailed Mann-Whitney tests were performed: P=0.0049 (between EGFP and eCB2.0 in saline) and 0.0581 (between EGFP and eCB2.0 with 2-AG).

f, Schematic diagram depicting the strategy for measuring G protein activation using the chimeric Gα_q-i protein.

g, Representative traces showing the jRGECO1a responses to 2-AG perfusion in cells expressing CB1R, CB1R+eCB2.0 or eCB2.0.

h, Dose-response curves of peak jRGECO1a ΔF/F₀ measured in cells expressing CB1R, CB1R+eCB2.0, or eCB2.0; n = 4, 4 and 3 cultures, mean ± s.e.m. Two concentrations of 2-AG were used for eCB2.0 expressed cells. Two-tailed Student’s t tests were performed: P=0.2392, 0.1455, 0.6711, 0.9191 and 0.8371 (between CB1R and CB1R+eCB2.0); P= 0.0156 and 0.0015 (between CB1R and eCB2.0).

i, G protein coupling was measured using a BRET G_βγ sensor in cells expressing CB1R, eCB2.0, or eCBmut; n = 3 experiments, mean ± s.e.m. Two-tailed Student’s t tests were performed: P=3.84E-05, 0.4082, 0.0699 and 0.2961.

j, β-arrestin coupling was measured using the Tango assay in cells expressing CB1R, eCB2.0, or eCBmut; n = 3 wells each, mean ± s.e.m.

k, Dose-response curves of eCB2.0 to 2-AG measured in cells expressing eCB2.0 or eCB2.0+CB1R; n = 3 wells each, mean ± s.e.m. Two-tailed Student’s t tests were performed: P=0.3036, 0.3231, 0.7697, 0.7900, 0.9723, 0.5482 and 0.1383.

***, p < 0.001; **, p < 0.01; *, p < 0.05; n.s., not significant.

Extended Data Fig. 4. Expression of GRAB_eCB2.0 has no significant effect on electrically evoked glutamate release in cultured neurons.

a, Fluorescence microscopy images of neurons expressing R^ncp-iGluSnFR (upper) and neurons co-expressing R^ncp-iGluSnFR and eCB2.0 (bottom). Similar results were observed for more than 20 neurons. Scale bar, 30 μm.

b, Example traces showing the electrical stimulation evoked glutamate signals.

c, Pseudocolor change in R^ncp-iGluSnFR fluorescence in neurons expressing R^ncp-iGluSnFR (upper) and co-expressing R^ncp-iGluSnFR and eCB2.0 (bottom) before and after the electrical stimulation. Shown are 25 regions of interest (ROIs) in one culture each.

d, Summary of peak R^ncp-iGluSnFR ΔF/F₀ measured in neurons expressing R^ncp-iGluSnFR (upper) or co-expressing R^ncp-iGluSnFR and eCB2.0 (bottom); n = 100 ROIs from 4 cultures each. Boxes show the first and third quartiles as well as the median (line), and the whiskers extend to the most extreme data point that is no more than 1.5× the interquartile range from the box. Two-tailed Mann-Whitney test was performed: P=0.2564.

n.s., not significant.

Extended Data Fig. 5. Estimated concentrations of electrically evoked 2-AG release in cultured neurons.

a, An example trace of ΔF/F₀ measured in eCB2.0 expressed neurons; the indicated concentrations of 2-AG were applied.

b, An example dose-response curve measured in neurons expressing eCB2.0. eCB2.0 signals evoked by 1 – 100 electrical pulses at 20 Hz and corresponding estimated 2-AG concentrations were indicated (green dots).

c, Summary of estimated 2-AG release concentrations evoked by 1 – 100 electrical pulses at 20 Hz; n = 3 cultures, mean ± s.e.m.

Extended Data Fig. 6. Spontaneous eCB transients in cultured neurons are sensitive to the CB1R neutral antagonist but not the action potential blocker.

a, Cumulative transient change in eCB2.0 fluorescence measured during 20 mins of recording in the absence (left) or presence (right) of 1 μM NESS0327. Pseudocolor images were calculated as the average temporal projection subtracted from the maximum temporal projection. Similar results were observed for 3 cultures. Scale bar, 100 μm.

b, Summary of the frequency of transient changes in eCB2.0 fluorescence measured in saline (Ctrl) and after NESS0327 application; n = 18 & 18 sessions from 3 cultures with 10-min recording/session. Boxes show the first and third quartiles as well as the median (line), and the whiskers extend to the most extreme data point that is no more than 1.5× the interquartile range from the box. Two-tailed Student’s t test was performed: P=1.68E-5.

c, Cumulative transient change in eCB2.0 fluorescence measured during 20 mins of recording in the absence (left) or presence (right) of 1 μM TTX. Pseudocolor images were calculated as the average temporal projection subtracted from the maximum temporal projection. Similar results were observed for 3 cultures. Scale bar, 100 μm.

d, Summary of the frequency of transient changes in eCB2.0 fluorescence measured in saline (Ctrl) and after TTX application; n = 12 & 14 sessions from 3 cultures with 10-min recording/session. Boxes show the first and third quartiles as well as the median (line), and the whiskers extend to the most extreme data point that is no more than 1.5× the interquartile range from the box. Two-tailed Student’s t test was performed: P=0.5972.

***, p < 0.001; n.s., not significant.

Extended Data Fig. 7. Detection of 2-AG, AEA, and DSI in GRAB_eCB2.0 expressed acute striatal slices.

a, Schematic diagram depicting the strategy for virus injection in DLS, followed by the preparation of acute brain slices used for electrical stimulation and recording.

b, Schematic diagram depicting the quantification of F₀ and decay time constant of the evoked eCB2.0 signal.

c, Quantification of relative F₀ and decay time constant of evoked eCB2.0 signals before and after JZL184 or URB597 treatment. n = 3 slices, mean ± s.e.m.

d, Fluorescence microscopy images of control and eCB2.0 expressed striatal slices. Recorded MSN neurons were loaded with Alexa 594. Similar results were observed for more than 10 neurons. Scale bar, 10 μm.

e, Depolarizing neurons in control and eCB2.0 expressed striatum caused similar depression on sIPSC. Three shadow regions correspond to baseline, early and late in c.

f, Summary of normalized charge recorded in MSNs in control and eCB2.0 expressed striatum during baseline, right after depolarization (early) and 16 s after depolarization (late); n = 8 and 13 neurons, mean ± s.e.m. Two-tailed Wilcoxon matched-pairs signed rant tests were performed: P=0.0078 (upper left), 0.0234 (upper right), 0.0134 (bottom left) and 0.0266 (bottom right).

**, p < 0.01; *, p < 0.05.

Extended Data Fig. 8. Detection of eCB signals in acute hippocampal and BLA slices using 2 photon imaging.

a, Schematic diagram depicting the strategy for virus injection in the hippocampal CA1 region, followed by the preparation of acute slices for electrical stimulation and 2-photon imaging.

b, (Left) fluorescence image of eCB2.0 expressed in the hippocampal CA1 region, showing the position of the stimulating electrode. (Right) pseudocolor images showing the change in eCB2.0 fluorescence at baseline and after 10 or 50 pulses applied at 20 Hz. The dashed circle shows the ROI for quantification. Similar results were observed for 5 slices. Scale bar, 100 μm.

c, Representative traces and summary of the peak change in eCB2.0 fluorescence evoked by electrical pulses applied at the indicated frequencies; n = 5 slices, mean ± s.e.m.

d, Time course of the change in eCB2.0 fluorescence; where indicated, AEA and AM251 were applied.

e, Representative traces of the change in eCB2.0 fluorescence evoked by electrical stimulation in the absence and presence of AM251.

f, Schematic diagram depicting the strategy for virus injection in the BLA region, followed by the preparation of acute slices for electrical stimulation and 2-photon imaging.

g, Pseudocolor images showing the change in eCB2.0 fluorescence after 20, 50 or 100 pulses applied at 20 Hz. Similar results were observed for 3 slices. Scale bar, 100 μm.

h, Traces of eCB2.0 fluorescence evoked by electrical pulses applied at the indicated frequencies.

i, Representative pseudocolor image, trace, and summary of peak change in eCB2.0 fluorescence upon 75 mM K⁺ ACSF perfusion. Scale bar, 100 μm. n = 3 slices, mean ± s.e.m.

Extended Data Fig. 9. Expression of GRAB_eCB sensors has minimal effect on animal behaviors.

a, Fluorescence images of coronal slices prepared from mice expressing GFP or GRAB_eCB2.0 in BLA. Similar results were observed for 6 mice. Scale bar, 1 mm.

b, Schematic diagrams showing the open field test (OFT) and the elevated plus maze test (EPMT).

c, Quantification of behavioral parameters in the OFT. n = 6 mice, mean ± s.e.m. Two-tailed Student’s t tests were performed: P=0.2084, 0.8737, 0.5858 and 0.4464.

d, Quantification of behavioral parameters in the EPMT. n = 6 mice, mean ± s.e.m. Two-tailed Student’s t tests were performed: P=0.2912, 0.5377, 0.6007, 0.3386, 0.3748, 0.4958 and 0.1411.

e, Schematic diagram showing the fear conditioning test.

f, Quantification of freezing behavior before, during and after conditioning. n = 6 mice, mean ± s.e.m. Two-way ANOVA test was performed: P=0.3799 (between two animal groups during conditioning); two-tailed Student’s t tests were performed: P=0.3297 and 0.8669 (during retrieval).

g, Quantification of averaged speed, running speed and averaged distance in control, eCB2.0 and eCBmut expressing mice; n = 19, 8 and 6 mice. Boxes show the first and third quartiles as well as the median (line), and the whiskers extend to the most extreme data point that is no more than 1.5× the interquartile range from the box. One-way ANOVA tests were performed: P=0.9017, 0.0681 and 0.4197.

n.s., not significant.

Extended Data Fig. 10. eCB and Ca²⁺ waves in mouse hippocampal CA1 region during seizure activity.

In vivo two-photon fluorescence images of eCB2.0 and jRGECO1a expressed in the mouse hippocampal CA1 region before and after stimulus evoked seizure activity. Frames were extracted from those shown in Supplementary Video 1. Seconds (s) after the stimulus are indicated. Similar results were observed for 6 mice. Scale bar, 100 μm.

Supplementary Material

Sup_Video1

Supplementary Video 1 eCB and calcium signals in mouse hippocampal CA1 during seizures ∣Fluorescence movies of eCB2.0 and jRGECO1a in the mouse hippocampal CA1 region during seizure activity, which is indicated by the LFP recording. The video is played at 3 times the speed. Similar results were observed for 6 mice.

DATA AVAILABILITY

Plasmids for expressing eCB2.0 and eCBmut used in this study have been deposited to Addgene (#164604–164612). Source data are provided with this paper.