A bistable inhibitory OptoGPCR for multiplexed optogenetic control of neural circuits

Information is transmitted between brain regions through the release of neurotransmitters from long-range projecting axons. Understanding how the activity of such long-range connections contributes to behavior requires efficient methods for reversibly manipulating their function. Chemogenetic and optogenetic tools, acting through endogenous G-protein coupled receptor (GPCRs) pathways, can be used to modulate synaptic transmission, but existing tools are limited in sensitivity, spatiotemporal precision, or spectral multiplexing capabilities. Here we systematically evaluated multiple bistable opsins for optogenetic applications and found that the Platynereis dumerilii ciliary opsin (PdCO) is an efficient, versatile, light-activated bistable GPCR that can suppress synaptic transmission in mammalian neurons with high temporal precision in-vivo. PdCO has superior biophysical properties that enable spectral multiplexing with other optogenetic actuators and reporters. We demonstrate that PdCO can be used to conduct reversible loss-of-function experiments in long-range projections of behaving animals, thereby enabling detailed synapse-specific functional circuit mapping.


Abstract
Information is transmitted between brain regions through the release of neurotransmitters from longrange projecting axons. Understanding how the activity of such long-range connections contributes to behavior requires efficient methods for reversibly manipulating their function. Chemogenetic and optogenetic tools, acting through endogenous G-protein coupled receptor (GPCRs) pathways, can be used to modulate synaptic transmission, but existing tools are limited in sensitivity, spatiotemporal precision, or spectral multiplexing capabilities. Here we systematically evaluated multiple bistable opsins for optogenetic applications and found that the Platynereis dumerilii ciliary opsin (PdCO) is an efficient, versatile, light-activated bistable GPCR that can suppress synaptic transmission in mammalian neurons with high temporal precision in-vivo. PdCO has superior biophysical properties that enable spectral multiplexing with other optogenetic actuators and reporters. We demonstrate that PdCO can be used to conduct reversible loss-of-function experiments in long-range projections of behaving animals, thereby enabling detailed synapse-specific functional circuit mapping.

Main
Even the simplest behaviors are coordinated by neural activity spanning multiple brain regions. Longrange projecting axons facilitate this by forming synaptic connections in distant circuits in the brain and periphery. Understanding the roles of these long-range connections in circuit activity and behavior requires techniques that allow selective manipulation of their function. While optogenetic tools allow the manipulation of neural firing with high temporal and spatial precision 1 , projection neurons often target several downstream regions via highly branched axonal collaterals 2,3 . Thus, manipulating the neuronal soma may result in a partial or even misleading picture of their contribution to circuit function. Instead, directly targeting the synaptic terminals of long-range connections can provide fine-grained insight into the role of specific neuronal pathways. However, direct suppression of synaptic terminal function poses unique challenges. Inhibitory optogenetic tools, such as the microbial light-driven ion pumps, have traditionally been used to silence synaptic transmission at axonal terminals [4][5][6][7][8] . Their inhibitory effect on synaptic release, however, is not only partial and short-lived, it can also induce unintended paradoxical effects, such as an increase of spontaneous transmitter release 9 . Chlorideconducting channelrhodopsins also proved unsuitable for synaptic silencing, as they depolarize the axon and can trigger antidromic firing due to the high chloride reversal potential in this subcellular compartment [9][10][11] . In contrast, targeting inhibitory GPCR pathways with bistable rhodopsins was shown to be effective for attenuating synaptic release in a projection-specific manner [12][13][14] .
We and others have recently shown that exogenously expressed light-activated animal rhodopsins can transiently inhibit synaptic transmission, by coupling to endogenous inhibitory G-proteins. While visual rhodopsins can be expressed in neurons and used to suppress synaptic release 12 , these photoreceptors can undergo bleaching (i.e. they lose their light-sensitive chromophore retinal), which reduces their efficacy under sustained illumination 15 . In contrast, bleaching-resistant non-visual rhodopsins have recently gained attention as light-activated tools for suppression of presynaptic release 16 . Like endogenous inhibitory GPCRs, these light-activated GPCRs (optoGPCRs) trigger the opening of G protein-coupled inwardly rectifying potassium (GIRK) channels, activate the Gα i/o signaling pathway (Fig. 1a) and efficiently suppress synaptic transmission through the inhibition of voltage-gated calcium channels (VGCCs) and the SNARE (soluble N-ethylmaleimide-sensitive-factor attachment receptor)-mediated fusion of synaptic vesicles with the presynaptic membrane 17,18 . Activated Gα i subunits can also reduce cyclic adenosine monophosphate (cAMP) production by adenylate cyclases (ACs), indirectly decreasing cAMP-dependent neurotransmission [19][20][21] .
Although progress has been made in developing inhibitory optoGPCRs, the existing tools are limited either in their spectral or temporal features. Two bistable rhodopsins, the trafficking-enhanced eOPN3 derived from Anopheles stephensi 13 (referred to herewith as AsOPN3) and the parapinopsin 14 from the Japanese lamprey Lethenteron camtschaticum (LcPPO), have recently been utilized as inhibitory optoGPCRs for presynaptic inhibition. The highly light-sensitive AsOPN3 has a broad action spectrum that spans the entire UV-visible range 13 . However, AsOPN3 activity cannot be rapidly reverted to the inactive (dark-adapted) state, and takes minutes to spontaneously recover to its non-signaling state 13,22 . LcPPO can undergo photoswitching between its active and inactive states by different wavelengths, thus allowing better temporal control 23,24 . However, LcPPO's limitation lies in its UV maximum activation wavelength (~370nm) and its broad inactivation spectrum 23,24 . These spectral properties restrict the wavelength range available for multiplexed applications with additional optogenetic actuators or fluorescence-based sensors. Especially for single-photon fiber photometry and miniature microscopy techniques, spectral multiplexing can be challenging with the current tools.
To improve and expand the capabilities of inhibitory optoGPCRs, we aimed for a new tool that retains the advantages of AsOPN3 and LcPPO but overcomes their limitations. We systematically screened a range of bistable opsins and evaluated their potential use as optoGPCRs based on their cellular biodistribution, spectral features and kinetic properties. Our screen revealed that the ciliary opsin 1 from Platynereis dumerilii (PdCO) 25,26 is a highly light-sensitive, bidirectionally switchable, versatile inhibitory optoGPCR. PdCO expresses well in mammalian neurons and allows robust, high-efficiency, and rapidly switchable presynaptic silencing across various cell types and preparations. With its unique spectral features, including a red-shifted activation wavelength and a narrow inactivation spectrum, PdCO is optimally suited for multiplexing with other optogenetic actuators and geneticallyencoded sensors.

Literature mining and functional benchmarking of optoGPCR candidates
We conducted a comprehensive literature search and identified a list of suitable optoGPCR candidates that could enable light-controlled inhibition of synaptic transmission. We collected information on retinal binding, spectral properties, switchability, G-protein coupling specificity, and activation of G-protein inward-rectifying potassium (GIRK) channels (Extended Data Fig. 1). Of the 32 described rhodopsins we selected for analysis, we identified 11 switchable variants that were most promising due to their coupling to Gα i/o or activation of GIRK channels. Including the AsOPN3 and LcPPO for comparison (Fig. 1b), we conducted a three-part benchmark to characterize the functional properties of these optoGPCRs. Our goals were to: 1) profile their G-protein coupling specificity ( Fig.  1d-f); 2) quantify GIRK channel currents induced by optoGPCRs activation (Fig. 1g-i); and 3) quantify their expression and membrane targeting in cultured neurons (Fig. 1j,k). Using the chimeric Gα bioluminescence assay (GsX 27 , Fig. 1d,e) we found that OlTMT1A, LcPPO, the PPOs from pufferfish (TrPPO1/2), PdCO, AsOPN3, and DrPPO1 couple to the inhibitory G i/o -protein family (G i/o/t/z , Fig. 1f). With the exception of TrPPO2 that displayed additional coupling to the G q/15 and G 12/13 , and BbOPN n= 6. f, Maximum bioluminescence response for light-activated optoGPCRs coupling to Gαsi (blue circles) and Gαso (green diamonds). n.d.: not determined, n= 6. g, HEK-cell experiments to measure optoGPCR-evoked GIRK currents with wholecell voltage-clamp recordings. h, Representative GIRK current traces recorded in HEK cells expressing the indicated optoGPCRs. Arrowheads and narrow bars indicate light application of 0.5 s, while wide bars indicate 10 s light activation (see Online Methods). i, Quantification of optoGPCR-evoked peak GIRK currents (n= 6 -16). j, Maximum intensity projection images from confocal stacks of neurons expressing the nine best-expressing optoGPCRs fused to mScarlet together with a cytosolic EYFP. Inverted grayscale images are gamma corrected (1.25). The merged projections are false color-coded by green (EYFP) and duo intense purple (mScarlet) lookup tables. k, Quantification of membrane expression index (blue circles) and the membrane fluorescence (green diamonds) of each optoGPCR, determined from equatorial z-slices (n = 10 -16). All data is shown as mean ± SEM. that showed non-selective G-protein coupling, we could not detect any G-protein activation other than for the G i/o pathway (Extended Data Fig. 2). Next, we expressed the optoGPCR candidates in human HEK293 cells that constitutively express GIRK2.1 channels, and measured the resulting G βγ -activated GIRK currents by whole-cell patchclamp electrophysiology (Fig. 1g,h). LcPPO, PdCO, TrPPO1, and AsOPN3 produced the largest GIRK current amplitudes (>700 pA), while the other variants induced currents smaller than 270 pA (Fig. 1i). BbOPN did not produce detectable light-induced currents, and GgOPN5L1 displayed a small inhibition of GIRK currents, which is consistent with its reported dark-activity 28 . We also measured optoGPCR-evoked GIRK activation in cultured neurons, where all optoGPCR except BbOPN and the OPN5 variants showed light-evoked GIRK conductance (Extended Data Fig. 4). To quantify the expression level and membrane targeting of the optoGPCRs, we co-expressed each of the opsins with a cell-filling EYFP fluorophore and used confocal microscopy to measure protein expression and membrane targeting ( Fig. 1j and Extended Data Fig. 3). In line with the electrophysiological recordings in HEK cells, LcPPO, PdCO, TrPPO1, and AsOPN3 showed the strongest expression and membrane targeting, and were only outperformed by DrPPO1 (Fig. 1k). We selected the seven best-performing variants (across all assays; Extended Data Fig. 4) and next tested their ability to attenuate synaptic transmission.

Benchmarking of bistable optoGPCRs in autaptic neurons
We expressed each of the selected optoGPCRs via rAAV2/1 transduction in autaptic neurons, a preparation in which single cultured neurons, grown on pre-seeded astrocyte micro-islands, form autaptic connections (Fig. 2a). Using whole-cell patch-clamp electrophysiology, we applied a series of paired pulses (pairs of depolarizations from -70 to 0 mV, every 5 s). We measured the induced excitatory postsynaptic current (EPSC) over three sequential periods: in the dark, post-lightapplication (one 500 ms pulse, 390 nm), and after administration of a 4.5 s-long 560 nm light pulse to inactivate the optoGPCR ( Fig. 2b and Extended Data Fig. 5). We applied the same protocol to nonexpressing controls, to correct for spontaneous EPSC rundown over time (Fig. 2c,d and Extended Data Fig. 5). Light activation of the PPOs from pufferfish (TrPPO1/2) and zebrafish (DrPPO1) had no effect on synaptic transmission, while OlTMT1A, PdCO, LcPPO and AsOPN3 significantly attenuated the AP-evoked EPSCs ( Fig. 2e and Extended Data Fig. 5). OlTMT1A and LcPPO could only attenuate transmission by 66±5% and 61±5%, respectively. Activation of PdCO and AsOPN3 yielded the strongest EPSC reduction, by 89±3% and 84±5%, respectively (Fig. 2e). As reported previously 13 , AsOPN3-mediated inhibition was long-lasting and could not be recovered by light application at different wavelengths. Green light-induced EPSC recovery was only partially possible for OlTMT1A, due to overlapping spectra of the opsin's dark-adapted and active states 29 . However, for PdCO and LcPPO, the green light pulse reliably induced recovery of synaptic transmission (Fig. 2e). traces (average of 7) pre (gray) and post illumination with different light pulse durations (blue and purple) recorded in the same autaptic neuron, for PdCO (top) and LcPPO (bottom). k, Quantification of release inhibition post illumination versus light flux for PdCO, LcPPO and AsOPN3, normalized to the inhibition for maximum light flux used. Solid lines show sigmoidal fits (n = 3 -17) l, Quantification of the absolute EPSC inhibition over 30 s post illumination at indicated wavelengths and maximum light flux for experiments as shown in j,k (n = 9 -17). Statistics: * indicates significance p<0.05; Kruskal-Wallis test followed by Dunn-Sidak multiple comparison; p(LcPPO, AsOPN3)= 1.82E-04. If not stated otherwise all data is shown as mean ± SEM.
In line with presynaptic inhibition, the frequency but not the amplitude of miniature EPSCs (mEPSCs) was reduced for the four optoGPCRs that showed light-induced EPSC attenuation (Extended Data Fig. 6). We observed an increased paired-pulse ratio following opsin activation, which was most strongly pronounced for PdCO and AsOPN3, providing further support for their presynaptic site of action (Extended Data Fig. 6).

Biophysical activation properties of PdCO
Given its promising performance in autaptic neurons and its photochromic properties, we characterized PdCO's biophysical properties in further detail in the context of synaptic inhibition and compared it with LcPPO and AsOPN3. First, we varied the wavelength of the activating 500 ms light pulse to generate action spectra for opsin activation (Fig. 2f). The wavelength needed for half maximum EPSC inhibition of PdCO was red-shifted by 40 nm compared to LcPPO ( Fig. 2g and Extended Data Fig. 7). In addition, synaptic transmission at this wavelength range was more effectively reduced by PdCO compared to LcPPO (Fig. 2h). PdCO activation with blue light (470 nm) showed transient inhibition that recovered with a time constant τ rec of 3.4±0.6 s (Fig. 2i). We next tested if continuous blue-light illumination of PdCO can evoke sustained inhibition that would similarly recover in the dark, without the need for a green pulse for inactivation. Indeed, continuous 470 nm illumination (2.83 mW/mm²) for 60 s reduced EPSCs by 85±1% in PdCO-expressing neurons. Evoked EPSCs recovered spontaneously after the cessation of light, with a time constant of 2.7±0.3 s. In contrast, we were not able to achieve inhibition with 470 nm light for LcPPO (Fig. 2i). Next, we varied the light pulse duration at the maximal effective wavelengths to compare the light sensitivity of PdCO, LcPPO and AsOPN3 (Fig. 2j). When quantifying the first EPSC after light activation, PdCO (EC 50 = 3.1±0.4 µW s/mm²) showed similar sensitivity to AsOPN3 (EC 50 = 1.9±0.3 µW s/mm², p=0.3217), whereas LcPPO showed lower sensitivity with an EC 50 of 30±2 µW s/mm² ( Fig.  2k and Extended Data Fig. 7). At the maximum pulse duration, AsOPN3 showed the strongest inhibition of (93±1%) followed by PdCO (82±3%) and LcPPO (67±4%; Fig. 2l).
We next tested whether PdCO can be activated with two-photon excitation in HEK293T cells coexpressing GIRK channels. To obtain the two-photon action spectrum for PdCO, we measured GIRK channel activation in cells expressing PdCO using whole-cell patch-clamp electrophysiology (Extended Data Fig. 7d). First, we applied raster scans at different wavelengths ranging from 700 nm to 1100 nm (3 mW, 20 s raster scanning) while applying a voltage ramp from -120 mV to +40 mV. Maximum GIRK channel activation was achieved with 800 nm at -120 mV (Extended Data Fig. 7e), in good agreement with one-photon activation. Next, we titrated the PdCO-coupled GIRK activation at 800 nm by varying light intensity. The half-maximum activation was determined to be 0.49±0.2 mW (Extended Data Fig. 7f). Data is normalized to bioluminescence reads pre-FSK application (n = 6). l, Normalized cAMP changes after light application, calculated by division of minimum response post illumination by maximum pre-illumination response of data as shown in k (n = 6). All data is shown as mean ± SEM.

G-protein specificity
During the electrophysiological recordings in autaptic neurons, we observed stronger and faster GIRK-mediated hyperpolarization in neurons expressing PdCO as compared to AsOPN3 or LcPPO (Fig 3a-d). To test whether the light-induced currents were caused by GIRK-channel coupling, we inhibited GIRK-channels with SCH23990 (Fig. 3a). Light-evoked GIRK-mediated currents were reduced by 77±6% after SCH23990 application in PdCO-expressing neurons, while inhibition of EPSCs by PdCO was not affected (Fig. 3e,f), indicating that synaptic inhibition via PdCO is independent of GIRK channel activity. Because PdCO showed very weak coupling to Gα o and Gα z in the GsX assay (Extended Data Fig. 2), we speculated that PdCO might have a different G-protein signaling bias that leads to the observed differences in GIRK activation. We therefore employed the TRUPATH assay 30 to characterize the G-protein signaling of LcPPO, AsOPN3 and PdCO in more detail (Fig. 3g). Both AsOPN3 and LcPPO showed long-lasting coupling to all members of the inhibitory G-protein family (Gα i-z ). In contrast, PdCO only coupled to Gα oA/B and Gα z , and not to Gα i1-3 ( Fig. 3h and Extended Data Fig. 8a). To exclude that inhibition of synaptic transmission is mediated by activation of Gα z , we inhibited Gα i and Gα o coupling with pertussis toxin (PTX). For all three opsins, PTX treatment abolished light-induced inhibition of EPSCs, indicating that Gα z does not contribute to presynaptic inhibition by these opsins (Fig. 3i and Extended Data Fig. 8b). As Gα i proteins are the main inhibitors of adenylate cyclases, we tested whether PdCO is capable of modulating cAMP production using a cAMP-dependent luciferase assay (GloSensor; Fig. 3j). As anticipated, PdCO activation did not have any detectable effect on cAMP production, whereas LcPPO and AsOPN3 activation led to a bioluminescence signal decrease of 63±1% and 62±1%, respectively (Fig. 3k,l).

Presynaptic inhibition in organotypic hippocampal slices
We next aimed to assay the inhibition efficacy of PdCO against LcPPO, the only other photoswitchable optoGPCR using organotypic hippocampal slice cultures. First, we confirmed that the biophysical properties of these two opsins were similar to those characterized in the autaptic culture preparation. Individual CA3 pyramidal neurons were transfected by single-cell electroporation to express either PdCO (Extended Data Fig. 9a) or LcPPO. We recorded GIRK-mediated currents evoked by light pulses at varying wavelengths and durations (Fig. 4a). The maximum GIRK current response for PdCO-expressing neurons was between 405 nm and 435 nm (Fig. 4b). Peak GIRK currents evoked by PdCO were higher than the ones induced by LcPPO at all tested wavelengths, even at a 10-fold lower light intensity for PdCO. Next, we varied the illumination time at the peak activation wavelengths of both optoGPCRs (365 nm for LcPPO and 405 nm for PdCO). PdCO-evoked GIRK currents showed maximum responses to light pulses with durations between 50 and 100 ms, and a higher amplitude than those evoked by LcPPO at the same pulse duration (Fig. 4c). We next activated the two optoGPCRs selectively at axonal terminals (Extended Data Fig. 9b) to compare their ability to suppress synaptic transmission. Presynaptic CA3 neurons were virally co-transduced with PdCO or LcPPO together with a soma-localized BIPOLES (somBIPOLES) 31 , to elicit red light (625 nm)-evoked APs in CA3 while avoiding potential cross-activation by PdCO illumination at CA1 (Fig.  4d). Red light pulses applied to the CA3 region reliably evoked EPSCs in CA1 cells, while application of 100 ms light pulse at 365 nm (10 mW/mm²) to LcPPO-expressing terminals reduced EPCS by 27±4% (Fig. 4e,f and Extended Data Fig. 9). Activation of PdCO with 10-fold lower light power at 405 nm led to a 78±5 % reduction in synaptic transmission, while no EPSC reduction was observed when somBIPOLES was expressed alone (Fig. 4e,f and Extended Data Fig. 9). For both optoGPCRs, attenuation of synaptic transmission was reliably recovered with 525 nm light (Fig. 4e,f). We next measured the stability of inhibition of synaptic release by PdCO, by stimulating PdCOexpressing Shaffer collaterals with a bipolar electrode at 0.1 Hz, while recording EPSCs in CA1 neurons (Fig. 4g). To exclude any somatic effects of the opsin and to avoid antidromic and recurrent activation of the CA3 network, we dissected out area CA3 prior to the recordings. Local application of a brief 500 ms light pulse in CA1 reduced evoked PSCs by 71 ± 0.3% and showed no spontaneous recovery over the time course of 25 minutes (Fig. 4h). This is in contrast to AsOPN3mediated inhibition, which spontaneously recovers with a time constant of approximately 5 minutes under identical experimental conditions 13 . In addition, we were able to recover transmission with 525 nm light and subsequently block synaptic transmission again with a second 405 nm pulse (Fig. 4h). Normalized EPSC amplitudes were not affected in non-expressing control cultures or non-illuminated PdCO cultures. (Extended Data Fig. 9).

Single-photon spectral multiplexing with PdCO
It is often informative to combine an optical readout of neuronal activity with optogenetic manipulations. For example, fiber photometry or, more recently, miniature microscopes can be combined with light stimulation at a different wavelength in the single-photon domain 32,33 . This requires spectral multiplexing of different optogenetic sensors and actuators, and benefits from minimizing spectral crosstalk 1,34 . To establish whether PdCO can be combined with red-shifted sensors or actuators, we next analyzed the wavelength-dependence of inactivation by varying the wavelength and irradiance of the inactivating pulse for both PdCO and LcPPO expressed in autaptic neurons. In these experiments, the optoGPCRs were activated at their peak excitation wavelength, and inactivation light was applied 30 s later (Fig. 5a). LcPPO showed a broad wavelength sensitivity that enabled near complete off-switching between 436 and 560 nm, while PdCO's inactivation sensitivity was maximal between 470 and 520 nm ( Fig. 5b and Extended Data Fig. 2b). We noted that the confined spectral window for inactivating PdCO might present an opportunity for spectral multiplexing with other optogenetic probes that are activated by longer wavelengths. We therefore titrated the light sensitivity for both optoGPCRs at 560 nm and determined that EPSC recovery at this wavelength is 6-fold more efficient for LcPPO (EC 50 = 61±2 µW/mm²) compared to PdCO (EC 50 = 372±163 µW/mm²; Fig. 5c), suggesting that PdCO is better suited for multiplexing applications with red-shifted sensors or actuators.
We next explored spectral multiplexing using the red-shifted calcium indicator FR-GECO1c 35 (Fig. 5d, top). PdCO, fused to EGFP for verification of expression, was co-expressed with FR-GECO1c in cultured neurons. In a tight-seal cell-attached patch-clamp configuration, we evoked APs at 0.2 Hz (Fig. 5d, bottom and Extended Data Fig. 10), resulting in reliable calcium transients in the FR-GECO1c signal (Fig. 5e, top). Blue light (445±10 nm) used to transiently activate PdCO, caused a 32±14% reduction in the amplitude of evoked calcium events (Fig. 5e,f). Notably, as reported for other GECO variants 36 , blue light application alone increased FR-GECO1c fluorescence, for which we corrected in our analysis (Extended Data Fig. 10). As GIRK activation can lead to reduced excitability or even suppression of AP firing, we blocked GIRK channels using SCH23390. Blue light application still decreased calcium transients by 18±4%, while in control cells only expressing FR-GECO1c, no reduction of calcium transients was detected (Fig 5e,f). Consistent with previous work 13,14 , this indicates that PdCO activation leads to the attenuation of somatodendritic VGCC activity.
Next, we combined PdCO with a soma-targeted variant of the red light-sensitive channelrhodopsin ChrimsonR 37 in a single bicistronic construct to allow the triggering of APs with red light, while simultaneously inhibiting synaptic transmission with the blue light-sensitive PdCO (Fig. 5 g-k). In cultured hippocampal neurons expressing this bicistronic construct, red light pulses (5 ms, 632 nm) generated photocurrents above 900 pA that reliably induced APs (Fig. 5 h,i). In non-expressing neurons, the same red light pulses caused reliable postsynaptic currents (PSCs) (Fig. 5h, bottom trace). When activating PdCO by a brief 390 nm light pulse (100 ms), repeated red light pulsing did not evoke PSCs anymore, indicating effective PdCO-mediated inhibition of synaptic transmission (Fig.  5j, upper left trace). Following green light application (512 nm) to recover transmission, red light-. CC-BY 4.0 International license made available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprint this version posted July 2, 2023. ; Figure 5: Single-photon spectral multiplexing of PdCO a, Example traces of experiments used to determine the spectral features and light sensitivity of optoGPCR inactivation in autaptic neurons. Samples were first illuminated with 390 nm (LcPPO) or 405 nm (PdCO) light for 500 ms, respectively, to inhibit EPSCs (black circles), followed by recovery with light delivery at the indicated wavelengths (equal photon flux density) for 4.5 s (colored traces) and finally completely recovered with 520 nm for at least 10s (gray traces). An average of 7 EPSC traces are shown, scaled to the fully recovered EPSC. b, Wavelength sensitivity of light-induced recovery. To correct for potential EPSC rundown, recovery z-scores were calculated using the mean of 4 EPSC post inhibition, prior recovery light at different wavelengths and the mean of 4 EPSC post full recovery with green light. n = 4 -7. c, Light titration of light induced recovery. Experiments were conducted as in b but at a fixed wavelength of 560 nm, while varying the light intensity between . CC-BY 4.0 International license made available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprint this version posted July 2, 2023. evoked PSCs were readily detectable again (Fig. 5j, lower trace); inhibition of evoked synaptic transmission could be achieved in a repetitive manner (Fig. 5j,k).

PdCO applications in cardiomyocytes and in vivo
To establish the efficacy of PdCO in modulating mouse behavior, we used it to unilaterally inhibit dopaminergic projections from the substantia nigra to the dorsomedial striatum, a neural pathway known to play an important role in animal locomotion 38 . We activated PdCO unilaterally in these axons during free locomotion and measured the resulting side bias, a behavioral test we previously established 13 . We expressed PdCO (or EYFP as control) bilaterally in substantia nigra pars compacta (SNc) dopaminergic neurons and implanted bilateral optical fibers above the dorsomedial striatum (DMS; Fig. 6a). Unilateral light activation caused an ipsiversive rotational bias in PdCO-expressing mice (Fig. 6b) that accumulated over time and ceased after illumination with green light (Fig. 6c.). This effect was consistent across PdCO expressing mice and absent in the EYFP-expressing control group (Fig. 6c,d).
To further test how PdCO inhibits specific synapses in vivo, we focused on locus coeruleus norepinephrine (LC-NE) modulation of pupil size [39][40][41] , which is largely mediated by disinhibition of the parasympathetic Edinger-Westphal (EW) nucleus 42 . LC neurons form vast, widespread projections that terminate in multiple brain regions 43 , making somatic inhibition highly non-specific. To selectively suppress LC axons terminating in the EW nucleus, we conditionally expressed PdCO unilaterally in NE neurons of the LC, and implanted optical fibers above the ipsilateral EW, and above the basal forebrain (BF, as control region; Fig 6e). Blue light application (447 nm) to EW led to robust dosedependent pupil constriction, whereas identical stimulation of BF did not (Fig. 6f-h). Notably, the ipsilateral pupil was significantly more affected by laser stimulation than the contralateral pupil. Given that pupil asymmetry does not occur under physiological conditions 44 , the observed lateralization provides strong evidence for successful pathway-specific inhibition (Fig. 6h).
. CC-BY 4.0 International license made available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprint this version posted July 2, 2023. ; Magenta and black colors depict ipsilateral or contralateral segments, respectively. Bottom: Average cumulative angle across PdCO-expressing (blue) and EYFP-expressing control mice (gray). Each mouse underwent two unilateral stimulations of each hemisphere, respectively, that was then averaged per mouse. n= 13-14 d, Quantification of the accumulated angle prior illumination (min. 9) compared to post UV-illumination (min. 20) for PdCO-expressing (blue, n=14) and EYFP-expressing control mice (gray, n= 13). The paired mean difference for both comparisons is shown as Cumming estimation plot. Each paired mean difference is plotted as a bootstrap sampling distribution. Mean differences are depicted as dots; 95% confidence intervals are indicated by the ends of the vertical error bars. Statistics: * indicates significance p<0.05; two-sided permutation t-test; 5.80E-03. e, Schematic of pupil experiment (see methods). f, Representative frames from pupil video recording at t=-5 (top),10 (middle) and 30 s (bottom) relative to 40 Hz laser stimulation onset. g, Mean pupillometry traces (bold lines, n=6) for 1 Hz (left), 10 Hz (middle), and 40 Hz (right) laser stimulation. Each time course denotes the median time-course of ipsilateral pupil diameter across trials in each subject in the basal forebrain (BF, gray) and the Edinger-Westphal nucleus (EW, blue). The vertical blue shaded area represents laser stimulation interval. Thin traces, individual mice; Thick traces, mean (n=6). h, Average pupil constriction (n=6 mice, matching the minimal value in time-courses shown in panel j) as a function of stimulation frequency (x-axis). Magenta dots and black diamonds represent EW stimulation effect on ipsilateral pupil or contralateral pupil, respectively. Bold dots/diamonds, average across mice (n=6).
. CC-BY 4.0 International license made available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprint this version posted July 2, 2023. ; https://doi.org/10.1101/2023.07.01.547328 doi: bioRxiv preprint Error bars, SEM across mice. Light dots, average of trials in each mouse. Note that EW stimulation is associated with dosedependent pupil constriction, and this is stronger for ipsilateral pupils. Statistics: * indicates significance p<0.05; multiple way ANOVA; p(placement)= 1.22E-20, p(frequency)= 9.62E-12, p(eye laterality)= 1.98E-04, p(placementfrequency)= 5.42E-10, p(placementeye laterality)= 8.71E-03. i, Top: representative beating trace (one dash per beat) of an atrial cardiomyocyte expressing PdCO, activated with different UV-light intensities and inactivated with green light. Bottom: Analysis of beats per minute (BPM) over time. j, Quantification of normalized beating frequency vs applied UV-light intensities from n = x different cells. Solid line depicts dose-response fit used for EC50 estimation. k, Quantification of beating reduction at 81 µW/mm 2 UV-light compared to non-illuminated conditions. Statistics: * indicates significance p<0.05; twotailed paired t-test; p= 5.35E-05. If not stated otherwise all data is shown as mean ± SEM.
Finally, we investigated the potential application of PdCO in non-neuronal cardiac tissue. In the heart, multiple GPCR-mediated mechanisms modulate pacemaking activity. Sympathetic stimulation is mediated via β-adrenergic Gα s -signaling while parasympathetic stimulation occurs through M2muscarinic receptor Gα i -signaling. Activation of Gα s or Gα i increases or decreases AC activity and cAMP levels, respectively. Elevated cAMP levels activate hyperpolarization-activated cyclic nucleotide-gated (HCN) ion channels leading to faster depolarization and beating frequency. While PdCO does not inhibit AC (Fig. 3j-l), it strongly couples to GIRK channels (Fig. 1g-i) and suppresses calcium transients (Fig. 5d-f). Since atrial cardiomyocytes endogenously express high levels of GIRK channels and rely on calcium influx for their electrical activity, we expressed PdCO in spontaneous active neonatal atrial cardiomyocytes and explored light-mediated suppression of pacing. We found that the spontaneous beating of PdCO-expressing cardiomyocytes could be suppressed after a brief violet light pulse in a light intensity-dependent manner (Fig. 6i,j). Inhibition of beating occurred with a half-maximal light dose of 2.8±0.4 µW/mm² (Fig. 6j), consistent with similar measurements in neurons. At 81 µW/mm², beating could be efficiently reduced by 80±4% (Fig. 6k). These results suggest that PdCO can be efficiently utilized in non-neuronal tissues with comparable efficiency and light sensitivity to neuronal preparations.

Discussion
Efficient presynaptic inhibition offers unique opportunities to study projection-specific contributions to behavior. While most previous optogenetic tools are not suitable for this approach 8,9,16,45 , two recently described optoGPCRs (AsOPN3 and LcPPO) could sufficiently provide axonal inhibition. However, each of these powerful tools has critical limitations that impair their utility for optogenetic applications, including a wide action spectrum and long spontaneous decay kinetics (AsOPN3) and a broad and light-sensitive deactivation cross-section that limits multiplexing applications (LcPPO). We therefore systematically compared a set of bistable type II rhodopsins with the goal of identifying optoGPCR candidates suitable for efficient presynaptic inhibition with precise temporal control and spectral properties suitable for optical multiplexing.
Both LcPPO and PdCO are photoswitchable optoGPCRs that can be activated with UV to blue light and inactivated with green light. For experiments requiring rapid switchability, these opsins have an inherent advantage over AsOPN3, which is activated more broadly with any wavelength between 390 and 640 nm, and only recovers through a thermal back-reaction in the dark. Thus, PdCO and LcPPO can provide superior temporal control over termination of optoGPCR signaling. PdCO showed faster onset compared to AsOPN3 and LcPPO, allowing a more precise timing of G-protein signaling onset. However, the rate limiting steps in signaling kinetics will be determined in all cases by the availability and mobility of the Gα and Gβγ subunits. Importantly, PdCO showed more robust suppression of presynaptic transmission when compared to LcPPO in hippocampal neurons, and its action spectrum is red-shifted by 40nm, allowing maximal activation with wavelengths up to 420nm. Blue light . CC-BY 4.0 International license made available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprint this version posted July 2, 2023. ; https://doi.org/10.1101/2023.07.01.547328 doi: bioRxiv preprint (>440nm) can be used to transiently activate PdCO since it is absorbed by both the inactive and active states and therefore drives both the forward-and back-reactions simultaneously.
When tested as a purified protein, the absorption maximum of PdCO (383nm) 26 is only shifted by 13nm compared to that of LcPPO (370nm) 23 . However, we were able to activate PdCO with wavelengths up to 470nm. For LcPPO, synaptic inhibition was only feasible up to 405nm in our recordings, and GIRK activation in organotypic slices dropped strongly above 405nm, in line with previous reports 24 . PdCO's red-shifted activation spectra can be explained by a low absorption crosssection of the active state that shifts the equilibrium wavelength between activation and light-induced recovery towards the red spectrum. The lower probability of light-induced inactivation is also reflected in the inactivation spectrum of PdCO, which is narrower compared to that of LcPPO. Intriguingly, LcPPO has been reported to be activatable with blue light (around 470 nm) in various experimental settings 14 . However, we were not able to achieve true blue light activation of LcPPO. Instead, we efficiently inhibited LcPPO activity in our experiments using 470 nm as reported previously 24 . This discrepancy might result from bandwidth limited light in our experiments, which eliminated lowwavelength photons.
We observed that PdCO activation led to transient synaptic inhibition when low light intensities at 405 nm or more red-shifted activation light is used (e.g. 445-470 nm). This indicates that if a low number of the activated G-proteins are recruited, the inhibitory effects can be transient, consistent with similar effects in chemogenetic actuators 46 . Although we demonstrated synaptic inhibition over 20 mins in organotypic slice preparations, care should be taken when using PdCO for long-lasting synaptic silencing experiments following only a brief single light pulse activation. Especially for in vivo experiments, if light delivery and expression levels are below saturation, PdCO-mediated inhibition could be short-lived; this can be overcome by repetitive light application or increased opsin expression. Furthermore, optoGPCR kinetics might vary between cell types, availability of heterotrimeric G-protein subunits and effectors/targets 47 , and input specific AP frequency and membrane depolarization [48][49][50] . Therefore, the inhibitory effect of presynaptic optoGPCRs should be tested by recording postsynaptic input reduction over time as discussed elsewhere 16 . Such experiments would be facilitated by the bicistronic constructs described above (Fig. 5g-k), which would allow co-expression of the red-shifted ChrimsonR with PdCO in the same neurons.
In our experiments, as well as across the optoGPCR literature, testing the same optoGPCR with different established assays of GPCR signaling can lead to vastly different outcomes. OPN5 homologs, for example, which did not couple to either Gα q or Gα i/o in the GsX assay. still generated a GIRK response. These optoGPCRs have recently been described to mediate Gα q coupling in various settings 51,52 . For PdCO, we observed efficient GIRK coupling as shown previously 26 but could only demonstrate very weak Gα o coupling in the GsX assay. The TRUPATH assay, however, revealed that PdCO selectively couples to Gα o but not to Gα i, which was confirmed by the demonstration that PdCO activation does not inhibit cAMP production. Nevertheless, we found that PdCO allows efficient silencing of presynaptic transmission, indicating that selective activation of the Gα o pathway can strongly suppress presynaptic release in all preparations tested in our study. The lack of PdCO impact on AC activity and the absence of effects on presynaptic cAMP can offer potential benefits as cAMP is involved in various intracellular processes such as proliferation, differentiation, survival, long-term synaptic potentiation, neurogenesis, and neuronal plasticity. However, potential modulation of cAMP by PdCO activation should not be completely excluded as a variety of ACs have been reported to be affected by different Gα and Gβγ subunits including Gα o 55 . It has also been shown that PdCO can transiently recruit Gα i under long lasting continuous and/or high-intensity illumination 53,54 , potentially by depleting available Gα o over time and therefore generating a signaling bias towards Gα i . While our primary focus in this study has been to develop new optoGPCRs for presynaptic inhibition, PdCO and many other optoGPCRs showed strong coupling to GIRK channels at the neuronal soma. Since PdCO showed both faster and stronger GIRK coupling in autaptic neurons compared to AsOPN3 and LcPPO, it could be used as a tool to reduce neuronal excitability when activated at the soma. However, not all neurons express GIRK channels and somatic inhibition might be absent in some cell types (e.g. medium spiny neurons in the striatum). Thus, when somatic inhibition is desired, anion-or potassium-conducting channelrhodopsins 10,55,56 might be more suitable, due to their strong inhibitory photocurrents and their millisecond-scale decay kinetics upon light offset. Nonetheless, by blocking GIRK channel activity we demonstrated that PdCO-mediated synaptic attenuation of transmission is independent of GIRK activity and can therefore be applied in neurons lacking these channels.
The diversity of genetically-encoded actuators and sensors provides a wealth of opportunities for multiplexed experiments, combining two or more such tools in a single experimental setting. As the activation spectrum of AsOPN3 covers the entire UV-visible range and due to its high light sensitivity, it requires careful handling and cannot be combined with other optical approaches apart from twophoton imaging 13 . In contrast, LcPPO and PdCO are both activated on the high energy visible spectrum (UV to blue light) and therefore do not bear the risk of cross-activation by other wavelengths used for imaging or optogenetic control. The narrow action spectrum of PdCO's light-induced backreaction to the inactive state is an attractive property for multiplexing with genetically-encoded tools that have red-shifted excitation spectra. Whereas one photon multiplexing has been demonstrated to be similarly possible with LcPPO 14 , we found that application of cyan to red light can cause a stronger inactivation of LcPPO compared to PdCO (Fig. 5a-c). For activation of larger brain areas, AsOPN3 might serve as a more suitable inhibitory optoGPCR due to its red-shifted activation spectrum and high light sensitivity. However, independent AsOPN3 activation at different brain loci might be less feasible due to potential cross-excitation by scattered photons. In this case, LcPPO and PdCO could serve as an alternative as UV-blue light is more effectively attenuated in neuronal tissues. Since AsOPN3 can also be activated by UV to blue light, these wavelengths can be used to excite AsOPN3 in settings where slow kinetics are desirable and activation by scattered light is a concern.
Taken together, we demonstrated that PdCO is a rapid, reversible, and versatile optoGPCR that mediates inhibition of synaptic transmission efficiently in diverse cell types in vitro and in vivo. With activation time constants in the sub 100-ms and inactivation switching times <10 s, PdCO serves as a fast inhibitory optoGPCR for precise presynaptic inhibition that expands and complements the collection of established and recently developed presynaptic optogenetic tools 16,57 . For manipulating the presynapse, PdCO could potentially serve as a suitable template to create optoGPCR chimeras with altered signaling specificity by exchanging the intracellular GPCR interface as previously demonstrated for other rhodopsin GPCRs 58-68 . PdCO's biophysical properties are highly suitable for one-photon spectral multiplexing approaches, which are becoming more common in the systems neuroscience field. We believe that PdCO, along with existing optogenetic sensors and with future improved, red-shifted indicators of neuronal activity, will serve as a valuable tool that will allow a better understanding of long-range neural communication in the brain.

Molecular biology and DNA constructs
Mammalian codon optimized Genes encoding optoGPCRs were synthetized (Twist biosciences, USA; except for LcPPO which was generously provided by P. Hegemann, Humboldt-Universität zu Berlin) and fused to a C-terminal rhodopsin 1D4 tag (TETSQVAPA). All genes were further subcloned . CC-BY 4.0 International license made available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprint this version posted July 2, 2023. ; https://doi.org/10.1101/2023.07.01.547328 doi: bioRxiv preprint in-frame with a C-terminal mScarlet into pcDNA3.1 vector under a CMV promoter or into pAAV vector under the CaMKIIα minimal promotor (CaMKIIα 0.4). Expression vectors additionally contained the Kir2.1 membrane trafficking signal (KSRITSEGEYIPLDQIDINV) and Kir2.1 ER export signal (FCYENEV) 69 , N-and C-terminal to mScarlet, respectively, as previously reported for AsOPN3 13 . The following genes were used for expression (NCBI GeneBank identifier; modifications if applied): GgOpn5l1 70 (AB368181; modified N-and C-termini originating from Xenopus tropicalis OPN5m to improve expression as reported elsewhere 28

Recombinant AAV vector production
For production of rAAV particles, human embryonic kidney cells (HEK293T) cells were seeded at 30±5% confluency and transfected 1d post seeding with plasmids encoding for AAV rep, caps of AAV2 and AAV1 or AAV9 and a vector plasmid encoding for rAAV cassette expressing the abovedescribed optoGPCRs using the PEI method 78 . 72h post transfection, cells were harvested and concentrated by centrifugation at 300g. The resulting cell pellet was resuspended in lysis solution (in mM): 150 NaCl, 50 Tris-HCl; pH 8.5 with NaOH). Cell lysis was performed by three freeze-thaw cycles and treated with 250 U/ml lysate benzonase (Sigma) at 37°C for 1.5h to remove genomic and unpacked DNA, followed by centrifugation at 3000 g for 15 min. Crude virus used for transducing neuronal cultures was filtered with sterile 0.45 µm PVDF filters (Millex-HV, Merck). To produce purified rAAVs, the virus-containing supernatant (crude rAAV) was purified using heparin-agarose columns, eluted with 0.5M NaCl and washed with phosphate buffered saline (PBS). The resulting viral . CC-BY 4.0 International license made available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprint this version posted July 2, 2023. ; https://doi.org/10.1101/2023.07.01.547328 doi: bioRxiv preprint suspension was concentrated with 100 kDa Ultra-15 centrifugal filters (Amicon), aliquoted, and stored at -80°C. Viral titers were quantified by real-time PCR.

optoGPCR mediated GIRK current recordings from human embryonic kidney cells
For the initial comparison of optoGPCR-evoked GIRK currents, optoGPCRs were transiently expressed in HEK293 cells stably expressing GIRK1/2 subunits (kindly provided by Dr. A. Tinker, Queen Mary's School of Medicine and Dentistry). Briefly, cells were maintained at 37°C, 5% CO 2 in high glucose DMEM supplemented with Geniticin (G418, GIBCO), 10% fetal bovine serum (FBS, Biological Industries) and penicillin/streptomycin (100 U/ml) and seeded onto poly-D-lysine coated . CC-BY 4.0 International license made available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprint this version posted July 2, 2023. coverslips in 24-well plates (Corning) and were additionally supplemented with 1µM 9-cis retinal (Sigma). One day post-seeding, pcDNA3.1-CMV-optoGPCR-mScarlet plasmids were transiently transfected using FuGENE HD (Promega; 0.75 µl/well, plasmid DNA, 250 ng/well) in serum free DMEM (50 µl/well). Currents from HEK293 cells stably expressing GIRK were recorded under visual guidance using a slice scope II (Scientifica, UK) with an Olympus LUMPlanFL N 40x/0.80 W objective under IR-DIC. A Lumencor SpectraX light engine was used to identify expressing cells via mScarlet fluorescence and for light application to toggle optoGPCR activation. In the case of non-switchable or slow cycling optoGPCRs (AsOPN3, OlTMT1A, GgOPN5l1), expressing cells were identified first and patched only after an additional 25 min in darkness. HEK cells were perfused with extracellular solution (in mM): 20 NaCl, 120 KCl, 2 CaCl 2 , 1 MgCl 2 , 10 HEPES, pH 7.3 (KOH), 320 mOsm (with D-glucose). Glass microelectrodes (1.5-2.5 MΩ) were pulled from thin-walled glass capillaries and filled with (in mM): 5 NaCl, 40 KCl, 2 MgCl 2 , 10 HEPES, 100 K-aspartate, 5 MgATP, 0.1 Na 2 GTP, 2 EGTA, pH 7.3 (KOH), 300 mOsm (with D-glucose). GIRK currents were recorded in whole-cell voltage-clamp mode at a holding potential of -50 mV. A Multiclamp 700B amplifier and Digidata 1440A digitizer (both Molecular Devices) were used to control and acquire electrophysiological recordings. Data was acquired at 10 kHz and filtered at 3 kHz. The different optoGPCRs were activated with a 500 ms (5 s in case of GgOPN5l1) light pulse close to their reported activation maximum with 10 nm narrow bandpass filters (Edmund optics). Light intensities for each wavelength were calibrated to the same photon flux corresponding to 0.92 mW/mm² at 520nm. The following center wavelengths were used (in nm): 520 AsOPN3, 450 OlTMT1A, 473 TrPPO2, 450 BbOPN. 546 GgOPN5l1, and 394 for all other optoGPCRs. Light induced recovery was induced by application of a 10 s 568±10 nm light pulse. Experiments were performed at 22±1°C. Maximum GIRK current amplitudes were determined using Clampfit 10.7 (Molecular Devices). Light intensities were measured with a calibrated S170C power sensor (Thorlabs). For two-photon activation of PdCO, electrophysiological recordings were performed on HEK293T cells (HEK293T/17, ATCC, #CRL-1573) as described previously 14 . In brief, pcDNA3.1-CMV-PdCO-mScarlet was co-transfected (1:3) together with pCAG-GIRK2/1-myc 80 using Lipofectamine 2000 (Invitrogen) according to manufacturer instructions and supplemented with 1 µM 9-cis retinal (Sigma). GIRK currents evoked by activation of PdCO were recorded under visual guidance using a Fluoview FVMPE-RS multiphoton imaging system using an XLPLN25XWMP2 objective (both Olympus). The extracellular solution contained (in mM): 140 NaCl, 20 KCl, 0.5 CaCl 2 , 2 MgCl 2 , 10 Glucose, 10 HEPES, pH to 7.3 with NaOH, 315±5 mOsm. Cells were patched with microelectrodes pulled from thin-walled glass capillaries (1-5 MΩ) and filled with (in mM): 120 K-gluconate, 5 NaCl, 0.1 CaCl 2 , 2 MgCl 2 , 1.1 EGTA, 10 HEPES, 4 Na 2 ATP, 0.4 Na 2 GTP, 15 Na 2 -phosphocreatine, pH=7.28, 290 mOsm. Whole-cell voltage-clamp current recordings in response to 200 ms ramp depolarizations from -120 to 40mV every 2s with a holding potential of -40mV were amplified and digitized (20 kHz) using a HEKA EPC10 (filtered at 3 kHz). Whole-cell and pipette capacitance transients were minimized, and series resistance was compensated by 70%. Two-photon excitation for PdCO was carried out using MaiTai and Insight tunable Ti:Sapphire lasers (Spectra Physics). For photostimulation, cells were centered in a 90x90 pixel square (0.4792mm/pixel) and scanned at a speed of 8µs/pixel. The spectral characterization was performed using a 20 s two-photon stimulation at 700, 800, 900, 1000, or 1100nm at 3mW laser power. For each cell, 3 to 5 different wavelengths were applied randomly. For a dose-response titration at 800 nm, a 10 s stimulation was performed across 0.1, 0.3, 1, 3, and 10 mW from low to high intensity. Twophoton intensities were calibrated at the sample focal plane using a thermal power sensor (S175C, . CC-BY 4.0 International license made available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprint this version posted July 2, 2023. ; https://doi.org/10.1101/2023.07.01.547328 doi: bioRxiv preprint Thorlabs) and power meter (PM100D, Thorlabs). Data was analyzed using IgorPro (Wavemetrics) and NeuroMatic 81 using custom macros. For each voltage ramp, the GIRK-mediated inward current was averaged over 5 ms at -120 mV holding potential.

Primary dissociated hippocampal neuron culture and gene delivery
Primary cultured hippocampal neurons were isolated from CA1 and CA3 hippocampal regions of either sex P0 Sprague-Dawley rat pups (Envigo). Neurons were digested with 0.4 mg/ml papain (Worthington) and seeded on Matrigel (1:30, Corning) coated glass coverslips in 24-well plates at a density of 65,000 cells per well. Neurons were maintained in a 5% CO 2 humidified incubator in Neurobasal-A medium (Invitrogen) supplemented with 1.25% FBS, 4% B-27 supplement (GIBCO), and 2 mM Glutamax (GIBCO). For inhibition of glial overgrowth, 200 mM fluorodeoxyuridine (Sigma) was added at DIV 4 (days in vitro). For confocal imaging or initial electrophysiological recordings of opsin expressing cultured primary neurons, opsin and cell-filling EYFP or GIRK2.1 encoding plasmids were co-transfected at DIV5 using a modified Ca 2+ -phosphate method 82 . Briefly, the neuronal cultured medium of a 24 well plate was collected and replaced with 400 ml serum-free modified eagle medium (MEM, Thermo Fisher Scientific). 30 µl transfection mix (2 mg plasmid DNA and 250 mM CaCl 2 in HBS at pH 7.05) was added per well, and cells were incubated for 1h to allow for transfection. Neurons were washed twice with MEM, and the medium was changed back to the collected original medium. Cultured neurons were used between 14-17 DIV for experiments. For Ca 2+ imaging experiments, cultured neurons were co-transduced with pAAV_CaMKII(0.4)-PdCO-EGFP-WPRE and pAAV_CaMKII(0.4)-FR-GECO1c-WPRE at DIV 1. Experiments were carried out between 14-21 DIV. Autaptic primary hippocampal neuronal cultures on glia cell micro-islands were prepared from P0 mice (C57BL/6NHsd; Envigo, Cat#044) of either sex as previously described 83 . 300 µm diameter spots of growth permissive substrate consisting of 0.7 mg ml -1 collagen and 0.1 mg ml -1 poly-D-lysine were applied with a custom-made stamp on agarose-coated coverslips. First, astrocytes were seeded and were allowed to proliferate in Dulbecco's modified eagle medium (DMEM) supplemented with 10% fetal calf serum and 0.2% penicillin/streptomycin (Invitrogen) for one week to form glia microislands. After changing the medium to Neurobasal-A supplemented with 2% B27 and 0.2% penicillin/streptomycin, hippocampal neurons prepared from P0 mice were added at a density of 370 cells cm -2 . Neurons were transduced with rAAVs (1.5 10 8 VG per well, titer-matched for all constructs) at DIV 1 and were recorded between DIV 14 and DIV 21.

Confocal imaging, quantification membrane targeting and expression levels
Primary cultured hippocampal neurons (transfected as described above) co-expressing the different opsins (pAAV-CaMKII(0.4)-opsin-mScarlet) together with a cell filling EYFP (pAAV-CaMKII(0.4)-EYFP) were fixed and permeabilized 4 days post transfection using 4% PFA for 15 min, washed 3 times with PBS and stained for 3 min with DAPI (5 mg/mL solution diluted 1:30,000). Coverslips were mounted using PVA-DABCO (Sigma) and fluorescence images were acquired using a Zeiss LSM 700 confocal microscope equipped with a Plan-Apochromat 63x/1.40 Oil DIC objective. For quantification of opsin expression in the membrane and cytosol, respectively, binary masks for EYFP and mScarlet signals were generated using fixed thresholding in imageJ 87 on a single equatorial z-slice per expressing neuron, identified visually with help of the nuclear DAPI stain. Expression analysis was restricted to the somatodendritic region by manual selection. The EYFP mask was subtracted from the mScarlet mask to generate a mask that restricts the analysis to the membrane only. Subsequently, the average pixel intensity was measured for the defined regions of interest. The expression index was calculated by subtraction of the whole cell mScarlet signal by the EYFP signal, divided by the sum of both signals.

In vitro electrophysiology on neuronal samples
Qualitative measurements of optoGPCR functionality using primary cultured neurons co-expressing pAAV-CaMKII(0.4)-optoGPCR-mScarlet together with pcDNA3.1-CMV-GIRK2.1 transfected as described above, were performed using the same setup as described for measurements on stably expressing GIRK2/1 HEK293 cells. Neurons were patched with microelectrodes (3.0-4.5 mΩ), filled with (in mM): 2 NaCl, 4 KCl, 10 HEPES, 135 K-gluconate, 4 Na 2 ATP, 4 EGTA, 0.3 Na 2 GTP, 290 mOsm, pH adjusted to 7.3 with KOH. Electrophysiological recordings obtained under continuous perfusion in Tyrode's medium (in mM): 150 NaCl, 4 KCl, 2 MgCl 2 , 2 CaCl 2 , 10 D-glucose, 10 HEPES; 320 mOsm; pH adjusted to 7.35 with NaOH. EPSCs from autaptic primary neurons were recorded under visual guidance using an Olympus IX51 inverted microscope using an Olympus UPlanSApo 20x/0.75 UIS2 objective under far infrared light (>665 nm) widefield illumination. A CoolLED P4000 served as a light source to identify expressing cells and for light application to toggle optoGPCR activation. In the case of non-switchable optoGPCRs (AsOPN3 and OlTMT1A), electrophysiological recordings were performed first, and cells were investigated for expression post recordings. Acquired data was excluded in case cells were not expressing. Autaptic neurons were constantly perfused with extracellular solution (in mM): 140 NaCl, 2.4 KCl, 10 HEPES, 10 D-glucose, 2 CaCl 2 , and 4 MgCl 2 (pH adjusted to 7.3 with NaOH, 300 mOsm). Cells were patched with microelectrodes pulled from quartz glass capillaries (3-4 mΩ), filled with (in . CC-BY 4.0 International license made available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprint this version posted July 2, 2023. ; mM): 136 KCl, 17.8 HEPES, 1 EGTA, 0.6 MgCl 2 , 4 MgATP, 0.3 Na 2 GTP, 12 Na 2 phosphocreatine, 50 U/ml phosphocreatine kinase (300 mOsm); pH adjusted to 7.3 with KOH. A Multiclamp 700B (Molecular Devices) amplifier and NI USB-6343 digitizer (National Instruments) were used to control and acquire electrophysiological recordings and the application of light stimulation via WinWCP software (University of Strathclyde, Glasgow). Data was acquired at 10 kHz and filtered at 3 kHz. Cells were kept at -70 mV and series resistance and capacitance were compensated by 70%. EPSCs were elicited by a 1 ms depolarization to 0 mV (50 ms interstimulus interval, every 5 s) resulting in an unclamped axonal action potential causing neurotransmitter release. SCH23390 (Tocris) was locally applied with a perfusion system (AutoMate Scientific ValveLink8.2). Pertussis toxin (0.5 mg/ml) was applied to the cultures 24 h before the recordings. For the initial comparison of EPSC reduction efficacy, the different optoGPCRs were activated with a 0.5 s, 390±10 nm (FB390-10, Thorlabs) and potential recovery was induced by a 4.5 s, 560±10 nm (FB560-10, Thorlabs). For spectral sensitivity measurements, light from the CoolLED P4000 was filtered with narrow bandpass filters mounted on a FW212C filter wheel (Thorlabs  88 . Alternatively, a second BX 51WI microscope (Olympus) equipped with a Double IPA integrated patch amplifier controlled by SutterPatch software (Sutter Instrument) was used for electrophysiological measurements. Patch pipettes with a tip resistance of 3-5 MΩ were filled with (in mM): 135 K-gluconate, 4 MgCl 2 , 4 Na 2 -ATP, 0.4 Na-GTP, 10 Na 2 -phosphocreatine, 3 ascorbate, 0.2 EGTA, and 10 HEPES (pH 7.2). Artificial cerebrospinal fluid (ACSF) consisted of (in mM): 135 NaCl, 2.5 KCl, 4 CaCl 2 , 4 MgCl 2 , 10 Na-HEPES, 12.5 D-glucose, 1.25 NaH 2 PO 4 (pH 7.4). All experiments were performed at room temperature (21-23°C) except the extracellular field stimulation experiments, performed at 33±1 °C. For experiments measuring GIRK currents, synaptic blockers D-CPP-ene (10μM), NBQX (10μM), Picrotoxin (100μM) (Tocris, Bristol, UK) were added to the HEPES-buffered ASCF and patched optoGPCR-expressing CA3 neurons were held at -70 mV during the measurements. In synaptic stimulation experiments (optogenetic and electrical), postsynaptic non-transfected CA1 neurons were held at -60 mV while recording PSCs in voltage clamp mode. Access resistance of the recorded CA1 neuron was continuously monitored and recordings above 20 MΩ and/or with a drift > 30% were discarded. A 16-channel pE-4000 LED light engine (CoolLED, Andover, UK) was used for optogenetic stimulation of the optoGPCRs. Light intensity was measured in the object plane with a 1918 R power meter equipped with a calibrated 818 ST2 UV/D detector (Newport, Irvine CA) and divided by the illuminated field of the Olympus LUMPLFLN 60XW objective (0.134 mm2) or of the Olympus LUMPLFLN 40XW objective (0.322 mm 2 ). For presynaptic somBIPOLES stimulation we used a fiber-coupled LED (400 µm fiber, NA 0.39, M118L02, ThorLabs) to deliver 5 ms red light pulses at 625 nm. In extracellular electrical stimulation experiments, afferent Schaffer collateral axons were stimulated (0.2 ms, 20-70 µA every 10 s) with a monopolar glass electrode connected to a stimulus isolator (IS4 stimulator, Scientific Devices).
. CC-BY 4.0 International license made available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprint this version posted July 2, 2023. ; https://doi.org/10.1101/2023.07.01.547328 doi: bioRxiv preprint

One-photon spectral multiplexing in cultured neurons
For calcium imaging experiments paired with transient PdCO activation, hippocampal neuronal cultures were co-transduced with pAAV-hSyn1(0.4)-PdCO-EGFP and pAAV-CaMKII(0.4)-FR-GECO1c as described above and experiments were carried out between DIV 14-21. Imaging was performed with SliceScope II (Scientifica, UK) using an Olympus LUMPlanFL N 40x/0.80 W objective and data were acquired with an ORCA-Flash4.0 digital camera (C11440, Hamamatsu). Data acquisition was carried out with 4x4 binning, 80 ms exposure at 10 Hz and FR-GECO1c was excited at 586±10 nm with an irradiance of 0.52 mW/mm². To transiently activate PdCO, 445±10 nm light (6.37 mW/mm²) was applied at 10 Hz, 5 ms pulse width between the imaging intervals. Cells were constantly perfused with extracellular solution (in mM): 140 NaCl, 2.4 KCl, 10 HEPES, 10 D-glucose, 2 CaCl 2 , and 4 MgCl 2 (pH adjusted to 7.3 with NaOH, 300 mOsm). To evoke AP-induced calcium transients, cells were patched in the tight-seal, cell-attached configuration as described by Perkins 89 . Briefly, cells were patched with low resistance microelectrodes pulled from glass capillaries (1.5-2.5 mΩ), filled with 145 mM NaCl. A train of APs was evoked by a 40 ms square current injection of 0.6 to 1.0 µA at a frequency of 0.2 Hz through the intact membrane patch (>3 GΩ seal resistance). The change in the FR-GECO1c signal ΔF/F 0 (where F 0 is the median fluorescence signal and ΔF=F(t)-F 0 ) was calculated in imageJ 87 and corrected for blue light induced increase in the fluorescence signal (Extended Data Fig. 10). A Lumencor SpectraX light engine served as light source and light intensities were measured with a calibrated S170C power sensor (Thorlabs). The spectra of light used to excite PdCO and FR-GECO1c were measured with an Ocean QE pro spectrometer (Ocean Optics). Experiments using pAAV-CaMKII(0.4)-stChrimsonR-EGFP-P2A-PdCO-WPRE were performed on the setup described above using the same extracellular solution. Neurons were patched at -70mV with microelectrodes (3.0-4.5 mΩ), filled with (in mM): 2 NaCl, 4 KCl, 10 -108]. The light intensity for each wavelength was calibrated to the same photon flux corresponding to 0.92 mW/mm² at 520nm. Short light pulses were applied (2 ms) and photocurrents measured under TTX (1µM), CNQX (20µM), AP-5 (50µM), recorded in both directions per expressing cell (red to blue and blue to red) and then averaged. For postsynaptic recording in non-expressing cells, neurons were measured with 10µM SCH233900 (10µM) to eliminate the potential contribution of GIRK currents. The light from the Lumencor SpectraX light engine was filtered with bandpass filters and the following light intensities were used: 390±9nm (0.155 mW/mm 2 ) 512±12nm (0.327 mW/mm 2 ) 632±11nm (7.015 mW/mm 2 ). A Multiclamp 700B amplifier and Digidata 1440A digitizer (both Molecular Devices) were used to control and acquire electrophysiological recordings. Data was acquired at 10 kHz and filtered at 3 kHz. Light intensities were measured with a calibrated S170C power sensor (Thorlabs). and plated on fibronectin coated 24 well plates (µ-Plate ibidi). Video microscopy of beating atrial cardiac cells was performed 3-5 days post transfection in a imaging cell chamber with 5 % CO 2 , 80% humidity and 37°C on a Nikon Eclipse Ti2 microscope with a objective S Plan Fl LWD 20x 0.7 NA objective. Changes in spontaneous beating frequency were recorded under IR light using a CMOS camera (Grasshopper3, Teledyne FLIR) and analyzed online using custom designed software (LabView, National Instruments) as described before 90 . Light stimulation was performed with a LedHUB (Omicron-Laserage) equipped with 385 nm LED (attenuated with a 10 % neutral density filter) and a 555 nm LED. Illumination was controlled by a recording system (PowerLab 4/35 and LabChart8 Software, AD Instruments). The 385 nm light intensity was varied logarithmically from 0.27 µW/mm² to a maximum of 132.94 µW/mm² every 25 s with a duration of 500 ms. A 10 s 555 nm light pulse was applied before the 385 nm pulse and repeated every 15 s. Light intensities were measured with an S130A power meter sensor at the objective level (Thorlabs).

Pupillometry experiments
Mice were anesthetized with (0.7-1% isoflurane) and placed in a stereotaxic frame. Optic patch cords connected the implanted optic fiber ferrules to a 447 nm laser (Changchun New Industries Optoelectronics Technology Co., Ltd., China). Pupillometry was performed by illuminating the eyes with infrared light (VGAC) and by continuous video monitoring (10 frames/s) synchronized with laser stimulation. We examined the effects of different stimulation frequencies (1,5,10,20 and 40 Hz) in two separate locations (EW or forebrain) on ipsilateral or contralateral pupil size (relative to injection and fiber). Each stimulation trial was 10 seconds long, with 20 ms pulses and a long 2-minute inter-trial interval to allow pupils to return to baseline. Trial order was randomized for stimulation frequency and stimulation location within each experiment. Data analysis (quantification of pupil area in video frames) was performed as previous studies using custom Matlab code 91 . Briefly, we fit a circle to the pupil area in the video images and, for each trial separately, normalized pupil area relative to the 4s pre-trial baseline to compute percent change dynamics in the [−5 30] s interval around laser stimulation. Trials were then averaged for each animal, eye, and condition separately (~8 trials per condition) to generate time-courses as seen in Fig. 6j. Maximal pupil constriction (through in timecourse) was defined as the minimal value in this pupil time-course.
In vivo optogenetic silencing of the nigrostriatal pathway Following recovery, mice underwent a single 15-minute habituation session, to habituate to handling, bilateral patch cord attachment and the open field arena. In experimental sessions, we attached individual mice to a patch cord unilaterally to measure PdCO-induced bias in locomotion. We recorded the free locomotion in an open field arena (50×50×50 cm) continuously over 30 minutes under nearinfrared illumination. After a 10-minute baseline no-light period ("baseline"), we delivered 500 ms light pulses (400 nm, 10 mW at the fiber tip), at 0.4 Hz for 10 minutes ("PdCO activation"), to the right or left sides (on separate sessions). Then we administered 5 light pulses (530 nm, 4 sec duration at 0.1 Hz, 10 mW at the fiber tip; "PdCO recovery") followed by no-light administration for the rest of the 10 minutes recovery period. A MASTER-8 pulse generator (A.M.P.I.) triggered the activation of both wavelengths, delivered by STSI-Optogenetics-LED-Violet-Green system (Prizmatix). A near-infrared LED placed at the video ROI was used as a synchronization signal (100 ms pulse duration, 0.1 Hz, for the entire 30 minutes session). These digital signals were split and recorded using Open Ephys acquisition board and used to synchronize the video and LED operation. Offline video processing and mouse tracking was done using DeepLabCut 92 . Briefly, we trained DLC to detect 7 features on the mouse body (nose, head center, left and right ears, left and right hips, tail base), the 4 corners of the arena and the ON or OFF states and coordinates of the video synchronization LED. X-Y coordinates of each feature were then further processed to complete missing or noisy values (large and fast changes in X or Y) using linear interpolation (interp1) of data from neighboring frames. This was followed by a low pass filtering of the signals (malowess, with 50 points span and of linear order). Finally, a pixel to cm conversion was done based on the videodetected arena features and its physical measurements. A linear fit to the nose, head, center and tail features defined the mouse angle with respect to the south arena wall at each frame. Following its dynamics over the session, we identified direction shifts as a direction change in angle that exceeds 20 degrees and 1 second. To achieve a comparable measurement between right-and lefthemisphere sessions, we measured motion in the ipsilateral direction as positive and contralateral motion as negative from the cumulative track of angle. The net angle gain was calculated as the sum of ipsilateral and contralateral angles gained over each time bin (1-or 10-minute bins as indicated). Results from the left-and right hemisphere sessions of each mouse were averaged and then used for statistical comparison between the PdCO and control groups.

Data analysis, quantification, and statistics
Phylogenetic trees were generated with phylogeny.fr 93 . In vitro electrophysiological recordings in cultured cells were analyzed using Clampfit 10.7 (Molecular Devices) as well as IgorPro (Wavemetrics) and NeuroMatic 81 for two-photon experiments. Analysis of mEPSCs data was performed using Easy electrophysiology (v2.3.3b) with a 0.37 correlation cutoff and a 15pA amplitude threshold due to artificial noise created by series resistance compensation. G-protein coupling assays (TRUPATH, GsX, GloSensor) were analyzed in Microsoft Excel. Confocal imaging and calcium imaging data was analyzed in ImageJ 87 . Data from organotypic slice recordings was analyzed with MATLAB. Atrial cardiomyocyte beating was analyzed using LabView (National Instruments). In vivo experiments were analyzed using Matlab and DeepLabCut 92 . Statistical analysis was performed with MATLAB, Graphpad Prism 9 and estimation statistics were performed online 94 . Sample sizes were similar to those commonly used in the field and no statistical tests were run to predetermine sample size. Blinding was performed in autaptic benchmark experiments (Fig. 2e) and behavioral experiments silencing the nigrostriatal pathway. Randomization was performed in the nigrostriatal pathway silencing experiments, for biophysical characterization of optoGPCRs in autaptic neurons, and for two-photon characterization. Automated analysis was used whenever possible. For autaptic neuron recordings, cells were excluded from analysis if the first EPSC amplitude was below 100pA, and from the analysis of the paired-pulse ratio if optoGPCR activation completely abolished the first EPSC. Further, were cells excluded from analysis if the access resistance was above 20 MΩ or if the holding current exceeded 200 pA. For organotypic slice culture recordings cells were additionally excluded from analysis if a EPSC amplitude drift > 30% occurred. For in vivo electrophysiology ( Figure 6), recording sessions in which no units showed visual stimulus-evoked activity were excluded from the analysis.