Design and engineering of opioid biosensors

We replaced amino acids between R257 in the intracellular loop (ICL) 3 and R6.24 on the transmembrane domain (TM) 6 of the human κOR, S247 in ICL3 and K6.24 in TM6 of the human δOR, and S6.23 in TM6 and K6.24 in TM6 of human µOR, with a circularly permuted green fluorescent protein (cpGFP), to generate κLight, δLight and µLight sensors, respectively (Fig. 1a,b and Extended Data Fig. 1a,b). The dynamic range of each sensor was optimized by screening linker compositions. In total, the dynamic ranges of 698 κLight variants, 64 δLight variants and 233 µLight variants were examined in response to U50,488, met-enkephalin (ME) and DAMGO, respectively (Extended Data Fig. 1c). To promote excellent membrane localization, we fused a telencephalin (TlcnC) tag31 or endoplasmic reticulum (ER) export motif (FCYENEV)32 followed by a chain of GS linker and the proximal restriction and clustering (PRC) tag33 to the C terminus of κLight, δLight and µLight. We named these new variants κLight1.3, δLight1 and µLight1, respectively. In addition, we mutated D3.22 of κOR and D3.32 in δOR in the binding pockets to attenuate the ligand binding, which led to two control sensors κLight0 and δLight0.

Fig. 1: Development of the opioid sensors.

a,b, Simulated structure of κLight (a) and δLight (b). c,d, Representative images of four independent transient transfections of κLight1.3 (c) and δLight (d) in HEK293T cells and cultured hippocampal neurons. Heat map indicates SNR upon addition of DynA8 (100 μM) or ME (100 μM). Scale bar, 20 μm (cells) and 50 μm (neurons). e,f, In situ titration of κLight1.3 (e) and δLight (f)-expressing HEK293T cells respond to ligands in a concentration-dependent manner (DynA13, blue; β-endorphin, gray; ME, black). Error bars represent the s.e.m. The highlighted area corresponds to a concentration range from 1 pM to 10 nM or 100 pM to 100 nM. Dyn, dynorphin. g,h, Schild plot of κLight1.3 (g) and δLight (h) dose response with 100 nM, 1 μM and 10 μM of naloxone. i,j, Schild plot of κLight1.3 (i) and δLight (j) dose response with 100 nM, 1 μM and 10 μM of Nor-BNI. k,l, Schild plot of κLight1.3 (k) and δLight (l) dose response with 100 nM, 1 μM and 10 μM of ICI 174864. m,n, Schild plot of κLight1.3 (m) and δLight (n) dose response with 100 nM, 1 μM and 10 μM of CTAP. o, Combined Schild regression with Nor-BNI and naloxone on κLight1.3. p, Combined Schild regression with Nor-BNI, naloxone and ICI 174864 on δLight. e–o, n = 4. Error bars represent the s.e.m.

When transiently expressed in mammalian HEK293 cells and dissociated neuronal cultures, we observed excellent membrane expression of κLight1.3, δLight and µLight. All three sensors were activated by their endogenous receptor agonists (100 µM), dynorphin A1-8 (DynA8), ME and β-endorphin, respectively (signal-to-noise ratios (SNR) values for κLight1.3 (HEK) = 7.5 ± 0.45; κLight1.3 (neuron) = 5.6 ± 0.2; δLight (HEK) = 16 ± 0.62; δLight (neuron) = 8.9 ± 0.43; µLight (HEK) = 4.7 ± 0.26) (Fig. 1c,d and Extended Data Fig. 1d). The ligand-induced responses (κLight1.3 change in fluorescence (ΔF/F); neuron) = 151% ± 5.1%; δLight ΔF/F (neuron) = 123% ± 19.4%; µLight ΔF/F (neuron) = 19.6% ± 3.2%) were blocked by naloxone (1 mM), which is an antagonist for all three receptors (Extended Data Fig. 1e).

To eliminate response variability due to inconsistent expression level of sensors via transient transfection, we developed HEK293T cell lines stably expressing κLight1.3, δLight and µLight. Using these cell lines, we characterized the promiscuity of endogenous opioid peptides on activating sensors34. First, all three sensors have consistent excitation peak wavelengths at 495 nm and emission peaks at 515 nm (Extended Data Fig. 1f). Second, in situ titration showed that all three sensors can be activated by three distinct endogenous opioid peptides but with different potency and efficacy. κLight1.3 responded to dynorphin A1-13 (DynA13) with an apparent half maximal effective concentration (EC50) of 89.8 pM, which is three magnitudes higher than β-endorphin and ME. However, at higher concentrations (>10 µM), κLight1.3 displayed higher fluorescence changes to β-endorphin, followed by DynA13 and ME (ΔF/F (κLight - DynA13) = 93.6% ± 3.9%; ΔF/F (κLight - β-endorphin) = 126.9% ± 8.6%; ΔF/F (κLight - ME) = 80.3% ± 1.8%; Fig. 1e). δLight was activated by ME with an EC50 of 6.5 nM, which is two orders of magnitude greater than DynA13 and β-endorphin, and had higher fluorescence efficacy compared to these two peptides (ΔF/F (δLight - DynA13) = 232.6% ± 6.8%; ΔF/F (δLight - β-endorphin) = 147.9% ± 4.1%; ΔF/F (δLight - ME) = 246.1% ± 4.6%; Fig. 1f). In contrast, we did not observe apparent responses of control sensors when the binding pocket was ablated (κLight0 or δLight0; Extended Data Fig. 1g,h). To further examine the selectivity for κLight1.3 and δLight in the context of neurons, we infected dissociated hippocampal neurons with AAV9-hSyn-κLight1.3 and AAV9-hSyn-δLight, respectively. We performed in situ titration in dissociated hippocampal neurons in the same context as in the HEK293T stable cell line. Expectably, the selectivity of both sensors in neurons is consistent to that in HEK293T cells (Extended Data Fig. 1i,j). However, all three endogenous opioid peptides showed similar sensor potency and efficacy for µLight activation (Extended Data Fig. 2a), suggesting that further improvement and engineering are required. Together, at presumed physiological conditions (pM–100 nM), both κLight1.3 and δLight are selective and sensitive to endogenous opioid peptides.

Next, we sought to determine the selectivity of antagonists acting on κLight1.3 and δLight. By running the in situ titration in antagonist mode35 using the same HEK293T stable cell lines, we were able to determine the selectivity of antagonists acting on κLight1.3 and δLight. In addition to naloxone, we chose nor-binaltorphimine (Nor-BNI), ICI 174864 and CTAP, which selectively antagonize κOR, δOR and µOR, respectively. As expected, increasing the concentration of naloxone (100 nM to 10 µM) shifted the apparent EC50 to the right for DynA13 and ME for κLight1.3 and δLight, respectively: naloxone inhibited δLight with twofold greater affinity than κLight (p2A (δLight - naloxone) = 7.64, pA2 (κLight - naloxone) = 5.68). Nor-BNI displayed slightly higher affinity in blocking κLight than δLight (pA2 = 8.28 and 7.3, respectively; Fig. 1g–j,o,p). We did not observe apparent antagonism of κLight by ICI 174864, whereas it effectively inhibited activation of δLight by ME (pA2 δLight - ICI 174864 = 7.17; Fig. 1k,l,o,p). The µOR-selective antagonist CTAP did not affect the EC50 of DynA13 or ME in either κLight or δLight, respectively (Fig. 1m–p).

Selectivity and pharmacology of the opioid biosensors

We next used a low concentration (10 nM) of a broad panel of endogenous and synthetic ligands to evaluate their rank order of response for inducing sensor fluorescence. We found that known κOR-selective endogenous peptides induced significantly greater fluorescence changes at κLight compared to δOR-selective or µOR-selective ligands. Among the dynorphin peptides, the shorter-form dynorphin DynA8 induced lower activation of κLight compared to DynA13. Interestingly, nalfurafine, a synthetic κOR agonist, elicited an almost twofold greater fluorescence change compared to the dynorphins (Fig. 2a). For δLight cells, enkephalins and δOR-selective agonists elicited larger responses compared to other ligands; deltorphin I displayed similar efficacy as ME and LE for δLight activation (Fig. 2b). Endogenous opioid peptide agonists at µOR, including β-endorphin, endomorphin, metorphinamide and BAM18, displayed various efficacies for κLight1.3 and δLight activation, although to a much smaller extent compared to κOR-specific and δOR-specific peptides. Notably, U50,488 and U69,593 selectively activated κLight over δLight, while SNC80 and SNC162 activated δLight over κLight, confirming the sensors’ specificity to receptor-specific small-molecule agonists (Fig. 2a,b).

Fig. 2: Pharmacological characterization of κLight and δLight.

a, Normalized ΔF/F of κLight1.3 upon addition of the listed compounds (10 nM). ΔF/F of all compounds are normalized to DynA13 (DynA13: 1 ± 0.03, DynA17: 0.89 ± 0.08, DynA8: 0.61 ± 0.04, DynB9: 0.97 ± 0.03, β-neoendorphin: 0.26 ± 0.03, nalfurafine: 1.91 ± 0.09, U69,593: 0.12 ± 0.06, U50,488: 0.42 ± 0.03, ME: 0.18 ± 0.005, LE: 0.24 ± 0.02, deltorphin I: 0.26 ± 0.02, DPDPE: 0.19 ± 0.04, SNC162: 0.009 ± 0.02, SNC 80: 0.17 ± 0.008, β-endorphin: 0.21 ± 0.01, endomorphin I: 0.15 ± 0.05, metorphinamide: 0.41 ± 0.06, BAM18: 0.48 ± 0.03, DAMGO: 0.17 ± 0.03, morphine: 0.08 ± 0.02, fentanyl: 0.26 ± 0.02, oxycodone: 0.16 ± 0.04, methadone: 0.04 ± 0.03, buprenorphine: 0.27 ± 0.01; n = 4 wells each. ****P < 0.0001, one-way analysis of variance (ANOVA) compared to DynA13 with Sidak’s multiple-comparisons test). Error bars represent the s.e.m. b, Normalized ΔF/F of δLight upon addition of the listed compounds (10 nM). ΔF/F of all compounds were normalized to ME (ME: 1 ± 0.01, LE: 0.84 ± 0.01, deltorphin I: 1 ± 0.07, DPDPE: 0.15 ± 0.01, SNC162: 0.54 ± 0.02, SNC80: 0.42 ± 0.03, DynA13: 0.15 ± 0.01, DynA17: 0.12 ± 0.004, DynA8: 0.58 ± 0.03, DynB1-9: 0.53 ± 0.01, β-neoendorphin: 0.26 ± 0.01, nalfurafine: 0.24 ± 0.03, U69,593: 0.014 ± 0.014, U50,488: 0.009 ± 0.004, β-endorphin: −0.03 ± 0.004, endomorphin I: 0.12 ± 0.01, metorphinamide: 0.07 ± 0.02, BAM18: 0.23 ± 0.01, DAMGO: 0.2 ± 0.01, morphine: 0.12 ± 0.005, fentanyl: 0.25 ± 0.04, oxycodone: 0.11 ± 0.01, methadone: 0.06 ± 0.006, buprenorphine: 0.22 ± 0.003; n = 4 wells each. ****P < 0.0001, ***P = 0.0006, one-way ANOVA compared to DynA13 with Dunnett’s multiple-comparisons test). Error bars represent the s.e.m. c–e, log s-slope values (in nM−1) of κOR (c), δOR (d) and µOR (e)-specific ligands plotted in triangle plots (κLight, blue; δLight, green; µLight, magenta). Higher s-slope values are located on the outer triangle. Enk, enkephalin.

We then used radar plots to compare the proportionality constant (s-slope) of various receptor-selective ligands for activating each sensor (Fig. 2c–e and Extended Data Table 1). The s-slope is a constant that links the variables of dynamic range (ΔF/Fmax) and EC50 of a given sensor response to a drug, defined as ΔF/Fmax/EC50. It highlights both the efficacy and potency of drugs on sensor responses36. By plotting s-slope values of individual ligands on three sensors as a radar plot, we found that the long forms of dynorphin are more potent in activating κLight1.3 than the short forms, the latter of which displayed considerable activity at δLight as well. Both nalfurafine and U50,488 were selective for κLight1.3 (Fig. 2c). The enkephalins (both ME and Leu-Enk (LE)), as well as β-endorphin, were highly selective for δLight, whereas deltorphin I and DPDPE displayed similar s-slopes for κLight1.3 and δLight. Despite low efficacy at κLight1.3, the s-slope of SNC80 was slightly higher at κLight1.3 than that at δLight (Fig. 2d). Notably, µLight was insensitive to morphine, whereas the latter induced slight fluorescence increases at κLight1.3 and δLight. In contrast, methadone activated all three sensors with similar efficacy and potency. Buprenorphine activates all three sensors but showed higher potency for µLight and δLight. On the other hand, other µOR-selective synthetic drugs, including DAMGO, fentanyl and oxycodone, engaged µLight with higher s-slopes compared to κLight1.3 and δLight (Fig. 2e). Interestingly, oxycodone and buprenorphine suppressed, rather than enhanced, µLight fluorescence; thus, the s-slope was calculated using the absolute ΔF/Fmax (Extended Data Fig. 2b).

To determine whether the insertion of cpGFP perturbs the ligand binding properties of these receptor-based opioid sensors, we first assessed the binding profile of both sensors and their corresponding receptors, followed by examining the ability of κLight1.3 and δLight to engage G-protein and β-arrestin pathways coupled to κOR and δOR, respectively. We conducted a radioligand binding assay using cells expressing each sensor and a panel of ligands that includes several endogenous peptides16,37,38. For cells stably expressing µLight, endogenous opioid peptides displaced [3H] diprenorphine binding with nM IC50 except for metorphamide (µM IC50). Specific binding detected in the presence of these peptides ranged from 34% ± 2% for peptide F to 82% ± 2% with BAM18. In the case of synthetic agonists, we see that DAMGO and oxycodone have nM IC50 while morphine and fentanyl have µM IC50. Interestingly, in the case of fentanyl, we found that it exhibits nM IC50 in CHO cells stably expressing µORs (Extended Data Table 2). For cells stably expressing δLight, the endogenous opioid peptides and the synthetic agonists displaced [3H] diprenorphine binding with nM IC50 except for peptide E (µM IC50). Specific binding detected in the presence of the endogenous peptides ranged from 32% ± 3% for BAM18 to 77% ± 4% with ME (Extended Data Table 2). For cells stably expressing κLight1.3, the endogenous opioid peptides and the synthetic agonist U69,593 displaced [3H] diprenorphine binding with nM IC50. Specific binding detected in the presence of the endogenous peptides ranged from 10% ± 5% for ME to 76% ± 2% with DynB13 (Extended Data Table 2). We next compared binding parameters of sensors with those of receptors as previously reported under similar conditions16. Using s-slope analysis, we found that binding parameters of sensors and receptors correlated in all three cases (Extended Data Fig. 2c–e). A positive correlation, especially for κLight and δLight, suggested that both radioligand binding assay and fluorescence assay can report peptides’ efficaciousness similarly. Where µLight shows a negative correlation to µOR indicates that the dynamic range and affinity of µLight still needs improvement to reliably report binding profiles for endogenous peptides.

Besides comparing the sensors with radioligand binding, we assessed the signaling conductivity directly on κLight1.3 and δLight. By implementing NanoBiT assay, we measured luminescence values indicating the elevation of β-arrestin1 upon addition of DynA17 comparing between κLight1.3 and κOR, and addition of DADLE comparing between δLight and δOR. Unsurprisingly, the addition of cpGFP eliminated the β-arrestin1-recruiting capability of κLight1.3 and δLight (Extended Data Fig. 2f). On the other hand, we assessed the DynA17 inhibition of forskolin-induced cAMP elevation by applying the GloSensor assay onto κLight1.3 and κOR, and same paradigm for DADLE inhibition onto δLight and δOR. The result indicates that neither κLight1.3 nor δLight is able to reduce the ligand-induced elevation of cAMP signals (Extended Data Fig. 2g).

Together, these data suggest that the cpGFP insertion eliminated the signaling conductivity of the receptors and is not likely to perturb the binding pockets of the parent receptor. Our studies demonstrate that peptide binding to an opioid sensor triggers fluorescence changes that correlate with the binding of the peptide to the receptor, making the sensors serve as useful tools to quantify differences in ligand-driven conformational changes between peptides.

Imaging dynorphin diffusion in brain tissue with κLight

Photoactivatable or ‘caged’ synthetic variants of opioid NPs or photosensitive nanovesicles can be activated with millisecond precision using short flashes of light and have been optimized for spectrally orthogonal use with GFP-based probes39. The spatiotemporal scale over which NP volume transmission occurs in brain tissue has been determined by combining photoactivatable NPs or nanovesicles, electrophysiological recording or cell-based NP biosensors. We thus asked whether κLight can report opioid peptide volume transmission in brain tissue using photo-uncaging experiments.

To choose the most appropriate κLight variant that balances dynamic range and sensitivity, we first examined the responses and kinetics of various κLight variants using photoactivable DynA8 (CYD8)29. We injected AAV9-hSyn-κLight1.x (top κLight variants including 1.2a, 1.2b, 1.2c and 1.3) into the dorsal striatum (dStr) of C57 mouse pups (postnatal day (P)0–P3) and prepared the brain slices after 3 weeks of expression (Fig. 3a). On the day of imaging, CYD8 was circulated in the bath and photo-uncaged with a 50-ms flash of 355-nm laser light over an area of 3,800 µm2, while imaging the responses of κLight with a 473-nm LED within the same region (Fig. 3b). Among all the κLight variants tested (Extended Data Fig. 3a), κLight1.3 yielded the greatest response (ΔF/F = 11% ± 1.4%; Fig. 3c,d), followed by κLight1.2a (ΔF/F = 9.09% ± 0.81%), κLight1.2c (ΔF/F = 6.84% ± 0.65%) and κLight1.2b (ΔF/F = 5.1% ± 0.51%; Extended Data Fig. 3b,c). The uncaging response was completely blocked by the presence of naloxone (0.5% ± 0.1%; Fig. 3d), confirming that the fluorescence change is due to ligand-dependent sensor activation, as opposed to being an artifact of the ultraviolet (UV) light flash. While κLight1.3 had the greatest ΔF/F, we noticed that its response was slow to decay in comparison to κLight1.2a (tauoff - κLight1.3 = 202.1 s, tauoff - κLight1.2a = 179.7 s, tauoff - κLight1.2b = 246.1 s, tauoff - κLight1.2c = 165.0 s; Fig. 3c and Extended Data Fig. 3b), presumably due to the higher affinity for dynorphins that results in slower peptide dissociation (Extended Data Fig. 3d).

Fig. 3: κLight1.3 characterization in acute brain slices.

a, Schematics shows imaging of striatal acute brain slices and photo-uncaging CYD8 with a 355-nm laser. b, Time-lapse images (semitransparent gray circle shows the field of UV illumination). Scale bar, 50 μm. c, Response of κLight1.3 to CYD8 photo-uncaging (blue, n = 6 slices) in the absence and presence of naloxone (Nalo, 10 μM; black, n = 3 slices). Solid lines represent the mean, and the shaded areas represent the s.e.m. d, Quantification of the peak ΔF/F evoked by CYD8 photo-uncaging. κLight1.3 (blue); 11.1% ± 1.36%, n = 6, + naloxone (black); 0.51% ± 0.12%, n = 3, P = 0.0011, two-tailed unpaired t-test. e, Time course of κLight1.2a after CYD8 (5 μM) photo-uncaging in the dStr. The dashed circle indicates the site of UV illumination. Heat map indicates ΔF/F (%). Scale bar, 50 µm. f, Summary of experiments determining the apparent diffusion coefficient, n = 7 slices from 4 mice. D* = 1.439 ± 0.37 μm2 s−1. g, Schematics show local electrical stimulation of hippocampal slice with trains of 1-s, 50-Hz stimuli with a 0.5-s interstimulus interval. HPC, hippocampus; ISI, interstimulation interval. h, Representative image showing expression of κLight1.3a in CA3 (top; scale bar, 0.5 mm) and zoomed in to visualize the localization of localization κLight1.3a to the membranes of neuronal processes in the dentate gyrus (middle; scale bar, 50 μm). Representative two-photon field of view from 15 stimulations (bottom) indicating the averaged intensity across all frames and z-score of responses in the representative field of view; scale bar, 20 μm. i,j, Average κLight1.3a responses to various electrical stimulation in the absence (i) and presence (j) of antagonists, Nor-BNI (gray) and ICI 174864 (green). Solid lines represent the mean, and shaded areas represent the s.e.m. k, Bar graph summarizing the peak fluorescence response to each stimulation condition. 15 stim (n = 8 slices): 14.3% ± 2.4%, 10 stim (n = 7 slices): 6.62% ± 0.8%, 5 stim (n = 7 slices): 4.28% ± 0.6%, 1 stim (n = 7 slices): 2.12% ± 0.3%, Nor-BNI (100 µM, n = 3 slices): 1.57% ± 1.2%, ICI 174864 (100 µM, n = 3 slices): 6.44% ± 0.3%. Ordinary one-way ANOVA with Bonferroni’s multiple-comparisons test, individual conditions compared to 15 stim, 15 stim versus 10 stim: **P = 0.044, 15 stim versus 5 stim: ***P = 0.0001, 15 stim versus 1 stim: ****P < 0.0001, 15 stim versus Nor-BNI: ***P = 0.0002, 15 stim versus ICI 174864: not significant (NS), P = 0.0525. Error bars represent the s.e.m.

We next examined whether sensor expression might alter the ability of peptide ligands to engage endogenous opioid receptors. For this experiment, we used κLight1.2a, which exhibited faster decay kinetics than κLight1.3 upon DynA8 photorelease (Extended Data Fig. 3b), yet still produced a relatively large ΔF/F. Adeno-associated viruses (AAVs) encoding κLight1.2a or GFP control were injected into the hippocampus of C57 pups (P0–P3) and allowed to express for a minimum of 3 weeks before acute slices were prepared for electrophysiology (Extended Data Fig. 3e). Parvalbumin interneurons in the CA1 region of the hippocampus express µOR and δOR, which act presynaptically to suppress synaptic transmission40. Although DynA8’s primary target is κOR, it also binds to µOR and δOR (for example, Fig. 2b and Extended Data Table 1)41. This allowed us to ask whether the activation of µOR and δOR by DynA8 is altered by the expression of κLight1.2a. To assay opioid receptor function, we recorded inhibitory postsynaptic currents (IPSCs) in pyramidal cells, evoked with a stimulation protocol that favors µOR-sensitive and δOR-sensitive parvalbumin synapses40 (Extended Data Fig. 3f). Photorelease of DynA8 using 5-ms flashes of 355-nm light produced a rapid, power-dependent reduction in IPSC amplitude that reversed over the course of several minutes (Extended Data Fig. 3g,h). Compared to GFP control, κLight1.2a expression altered neither the degree of IPSC suppression, nor the time course of IPSC recovery in response to DynA8 photorelease across all light power densities examined (Extended Data Fig. 3i,j). These results suggest that κLight1.2a expression does not result in sufficient ligand buffering as to perturb the activation of endogenous opioid receptors.

We next measured the spread of DynA8 in space and time. AAV1-hSyn-κLight1.2a was injected into dStr and imaging was performed 3 weeks after injection (Fig. 3a). Small volumes of DynA8 were rapidly photoreleased using a focused 25-µm-diameter spot of 355-nm light (Fig. 3e) while monitoring sensor activation at distances of up to 125 µm away. We observed that the peak ΔF/F decreased with increased time from uncaging and with distance from the uncaging site (Extended Data Fig. 4a). For each video frame after uncaging, we plotted the fluorescence profile as a function of distance from the uncaging spot and extracted the ΔF/F half-width, which was used to compute an effective diffusion coefficient (D*) of 1.4 ± 0.4 µm2 s−1 (n = 7 slices from four mice) for DynA8 in dStr (Extended Data Fig. 4b–d). These results suggest that DynA8 can reach receptors over 100 µm away from release sites within several seconds of release in the dStr.

Two-photon imaging of dynorphin release via electrical stimulation

It has been historically difficult to determine the electrical parameters that can effectively trigger the release of endogenous opioid peptides in brain tissue. We thus examined if κLight can detect endogenous opioid peptide release triggered by electrical stimulation ex vivo. To do so, we first improved the basal fluorescence of κLight1.3 by integrating CYKIWRNFKGK as linker 1 and SVISKAKIRTV as linker 2 derived from the oxytocin sensor MTRIAOT42 (Extended Data Fig. 3a). This new variant, named κLight1.3a, displayed a similar dynamic range (κLight1.3 at 155% ± 11.6%, κLight1.3a at 152% ± 29.5%, P = 0.92, unpaired t-test), but >2× the basal brightness compared to κLight1.3 (κLight1.3 at 25 ± 0.08, κLight1.3a at 61.8 ± 7.6, P = 0.0075, unpaired t-test; Extended Data Fig. 4e,f). To validate that κLight1.3a retain the same selectivity, we performed in situ titration in dissociated hippocampal neuronal cultures using peptides DynA13, ME and β-endorphin. As expected, κLight1.3a showed high selectivity to DynA13 over the other two endogenous peptides (Extended Data Fig. 4g). Immunoreactivity studies have shown abundant dynorphin stored in dentate granule cells, dynorphin dynamics in CA3 have also been shown to have an association with stress under various behavior paradigms, and dynorphins have been shown to inhibit excitatory neurotransmission and prevent the induction of long-term potentiation in hippocampus43,44,45. We sparsely expressed κLight1.3a in CA3 by delivering AAV1-CAG-DIO-κLight1.3a in combination with AAV1-hSyn-Cre (Fig. 3g). After 3 weeks of expression, we observed bright labeling of neurons in CA3 and dentate gyrus with clear processes in the basal state and distribution of responses in the field of view using two-photon imaging (Fig. 3h).

Next, we evaluated the responses of κLight1.3a to a range of electrical stimuli parameters applied locally via a stimulating electrode in CA3. Trains of electrical stimuli (1 s at 50 Hz, 0.5-s interstimulus interval) produced sustained fluorescence increases that rapidly decayed upon cessation of the stimulus (Fig. 3i), with an increasing number of stimuli driving larger maximum fluorescence responses (15 stimulations: 14.3% ± 2.4%, 10 stimulations: 8.39% ± 1.9%, 5 stimulations: 4.28% ± 0.6%, 1 stimulation: 2.12% ± 3.3%; Fig. 3k).

The response to 15 stimuli was strongly attenuated by the addition of the κOR antagonist nor-BNI (100 μM, ∆F/F = 1.57% ± 1.2%), consistent with the observed fluorescence increase resulting from activation by endogenous peptide. In the presence of δOR antagonist ICI 174864 (100 μM), the responses were decreased but not statistically significant (Fig. 3j,k; ∆F/F = 6.44% ± 0.3%).

Probing the effect of receptor-selective opioid ligands in vivo

We next determined if κLight and δLight can be activated by systemic administration of exogenous small-molecule drugs in vivo. We injected AAV9-hSyn encoding κLight1.3 or δLight, and κLight0 or δLight0 in the arcuate nucleus (ARC) of the hypothalamus46, hippocampal CA3 region43 and NAc30, areas abundant in κOR and δOR. We next implanted fiber-optic ferrules above each injection site and recorded the fluorescence of κLight and δLight upon intraperitoneal (i.p.) injection of opioid receptor-selective ligands using fiber photometry (Fig. 4a and Extended Data Fig. 4h–j).

Fig. 4: In vivo drug pharmacology imaged with κLight and δLight.

a, Experimental schematics of κLight1.3 and δLight injection in the hypothalamus (ARC), the hippocampal CA3 and the NAc, followed by imaging with fiber photometry during drug injection. b, κLight1.3 response in ARC to different doses of U69,593, 3 mg per kg body weight (light blue), 1 mg per kg body weight (blue) and 3 mg per kg body weight U69,593 + 4 mg per kg body weight naloxone (black); n = 7 animals. Solid lines represent the mean, and the shaded area represents the s.e.m. Bar graph indicating the peak z-score of each response, 3 mg per kg body weight + naloxone: 0.4% ± 0.6%, 1 mg per kg body weight: 7.0% ± 1.9%, 3 mg per kg body weight: 15.9% ± 3.1%, ordinary one-way ANOVA with Tukey’s multiple-comparisons test, 1 versus 3 *P = 0.012, 1 versus Nalo *P = 0.029, 3 versus Nalo ****P < 0.0001. c, κLight1.3 response to different doses of U50,488 in CA3, 10 mg per kg body weight (light blue), 5 mg per kg body weight (blue) and 10 mg per kg body weight U50,488 + 10 mg per kg body weight naloxone (black) in CA3; n = 3 animals. Solid lines represent the mean, and shaded areas represent the s.e.m. Bar graph indicating the peak z-score of each response, 10 mg per kg body weight + naloxone: −2.9% ± 0.8%, 5 mg per kg body weight: 2.7% ± 1.8%, 10 mg per kg body weight: 11.1% ± 3.2%, ordinary one-way ANOVA with Dunnett’s multiple-comparisons test, **P = 0.0072. d,e, δLight response to different doses of SNC162 in ARC (d) and NAc (e), 5 mg per kg body weight (light green), 2.25 mg per kg body weight (green) and 5 mg per kg body weight SNC162 + 4 mg per kg body weight naloxone (black) in ARC and NAc; n = 4 animals. Solid lines represent the mean, and shaded areas represent the s.e.m. Bar graph indicating the peak z-score of each response; in ARC: 0.2% ± 0.7%, 2.25 mg per kg body weight: 2.4% ± 1.0%, 5 mg per kg body weight: 7.3% ± 2.4%, ordinary one-way ANOVA with Tukey’s multiple-comparisons test, *P = 0.0258; in NAc: 1.7 ± 0.1%, 5 mg per kg body weight: 7.5% ± 2.2%; two-tailed unpaired t-test, *P = 0.0185. In b–e, error bars represent the s.e.m.

In each case, we observed dose-dependent fluorescence increases in response to systemic drug i.p. treatment, which were blocked by the nonselective opioid receptor antagonist naloxone. In the ARC, κLight1.3 responded to the κOR-selective agonist U69,593 with a robust increase in fluorescence within a few minutes of drug injection (1 mg per kg body weight: z-scorepeak = 7.0 ± 1.9, 3 mg per kg body weight: z-scorepeak = 15.9 ± 3.05). Co-injection of naloxone (4 mg per kg body weight) drastically attenuated the response to U69,593 (3 mg per kg body weight; U69,593 + naloxone z-scorepeak = 0.39 ± 0.59; Fig. 4b). In CA3, the κOR-selective agonist U50,488 similarly activated κLight1.3 in a dose-dependent manner. Again, the response to U50,488 (10 mg per kg body weight) was completely blocked by co-injecting naloxone (10 mg per kg body weight; 5 mg per kg body weight: z-scorepeak = 2.68 ± 1.8; 10 mg per kg body weight: z-scorepeak = 11.1 ± 3.2; U50,488 + naloxone: z-scorepeak = −2.86 ± 0.83; Fig. 4c).

In the ARC, SNC162 administration produced increases in δLight fluorescence (2.25 mg per kg body weight: z-scorepeak = 2.4 ± 1.0; 5 mg per kg body weight: z-scorepeak = 7.28 ± 2.4) that were blocked by naloxone (4 mg per kg body weight) co-injected with SNC162 (5 mg per kg body weight; SNC162 + naloxone: z-scorepeak = 0.19 ± 0.72; Fig. 4d). In the NAc, the administration of SNC162 (5 mg per kg body weight) also increased δLight fluorescence (SNC162: z-scorepeak = 7.45 ± 2.20), and this was again blocked by naloxone (4 mg per kg body weight; SNC162 + naloxone: z-scorepeak = −1.66 ± 0.11; Fig. 4e).

Importantly, we did not observe fluorescence changes in response to agonist when the non-functional mutant sensors κLight0 or δLight0 were expressed in the ARC, CA3 and NAc (Extended Data Fig. 4k,l). These results suggest that both sensors can be faithfully activated by receptor-specific agonists in vivo and ensure a good dynamic range, adequate expression and fiber-expression alignments as a foundation for the following optogenetic and behavioral experiments.

Measuring dynorphin release via circuit-specific photostimulation

Although optogenetics has been broadly used to trigger neuromodulator release and neural activity, direct monitoring of peptide release triggered by optogenetic stimulation in vivo, especially in a circuit-specific manner with high temporal resolution, has not been measured optically. NAc contains abundant dynorphin, and previous studies have demonstrated that targeting the Dyn-κOR system in the nucleus accumbens shell (NAcSh) can modulate both rewarding and aversive behaviors47,48. Furthermore, previous work has demonstrated the ability to measure the optogenetically evoked release of dynorphin in the NAcSh using in vivo opto-dialysis30. Studies have also shown that the basolateral amygdala (BLA) sends dense, functional excitatory projections to the NAcSh and that these terminals are sensitive to modulation by Dyn-κOR49,50. We, therefore, set out to determine if κLight can detect photostimulated release in vivo in BLA to NAcSh projections.

To detect dynorphin signaling at κOR-expressing neurons, we injected κOR-Cre mice with AAV5-CAG-DIO-κLight1.3a and implanted optical fibers in the NAcSh. A subset of mice was also injected with the red-shifted opsin ChRimson (AAV5-DIO-EF1a-ChRimson-tdTomato) in the BLA (Fig. 5a–c and Extended Data Fig. 5a); ChRimson-lacking mice served as a negative control to determine if optical stimulation produced artifactual dynamics in κLight1.3a fluorescence. We first examined the response of κLight1.3a to the agonist U50,488 in these mice (Fig. 5d). U50,488 (10 mg per kg body weight; i.p.) administration resulted in a rapid, sustained and robust increase in the fluorescence of κLight1.3a. This increase was significantly attenuated when the animals were pretreated with the short-acting, reversible κOR antagonist JNJ-67953964 (ref. 51; aticaprant, 3 mg per kg body weight; i.p.; P = 0.034, paired t-test), demonstrating the selectivity of κLight1.3a responses in vivo (normalized peak, P = 0.0344, paired t-test, normalized area under the curve (AUC), P = 0.0138, paired t-test; Fig. 5e–h).

Fig. 5: Imaging optogenetically stimulated dynorphin release with κLight1.3a.

a, Schematic showing κLight1.3a-expressed NAcSh and ChRimson into the BLA of κOR-Cre+ mice. b, Representative ×20 coronal image (left) showing expression of κLight1.3a (green), ChRimson (red), DAPI (white) and fiber placement in the NAcSh (left; scale bar, 200 μm), and ChRimson (red) and DAPI (white) in the BLA (right; scale bar, 200 μm) from six animals, which showed similar results. c,d, Schematic of in vivo head-fixed stimulation-evoked dynorphin release (stimuli occurred at 0, 180 and 360 s; c) and agonist/antagonist drug injection (10 mg per kg body weight U50,488 and 3 mg per kg body weight aticaprant + 10 mg per kg body weight U50,488; d) experiments. e,f, Mean (e) and heat map (f) of κLight1.3a activity either averaged across all animals (e) or from individuals (f) following i.p. injections of vehicle (veh) + U50,488 (dark) and aticaprant + U50,488 (light; n = 6 animals). Solid lines represent the mean, and shaded areas represent the s.e.m. g,h, Normalized peak fluorescence (g) and AUC (h) of single trials during the injection period (0–50 min; U50: 1 ± 0.23, Atic + U50: 0.29 ± 0.07; two-tailed paired t-test, *P = 0.034, *P = 0.014, n = 6 animals). Data are represented as the mean ± s.e.