Selective control of synaptically-connected circuit elements by all-optical synapses

Understanding percepts, engrams and actions requires methods for selectively modulating synaptic communication between specific subsets of interconnected cells. Here, we develop an approach to control synaptically connected elements using bioluminescent light: Luciferase-generated light, originating from a presynaptic axon terminal, modulates an opsin in its postsynaptic target. Vesicular-localized luciferase is released into the synaptic cleft in response to presynaptic activity, creating a real-time Optical Synapse. Light production is under experimenter-control by introduction of the small molecule luciferin. Signal transmission across this optical synapse is temporally defined by the presence of both the luciferin and presynaptic activity. We validate synaptic Interluminescence by multi-electrode recording in cultured neurons and in mice in vivo. Interluminescence represents a powerful approach to achieve synapse-specific and activity-dependent circuit control in vivo.

A wealth of new tools are revolutionizing neuroscience by allowing direct control of specific subpopulations of neurons for brief and sustained time periods (e.g., opto-, chemo-and sonogenetics; [1][2][3]. The ability to regulate genetically identified neurons in selected brain areas has been a significant benefit to studying neural dynamics and its link to behavior. However, information processing leading to percepts, memories and/or actions requires multiple nodes acting in sequence within a network, with multiple cell types in multiple areas conducting specific cell-to-cell communication. The ideal tool(s) in the next generation of approaches will allow intersectional circuit dissection-specifying and regulating participants at multiple stations. Further, tools for fully understanding systems underlying behavior will allow them to demonstrate natural activity, enhancing or suppressing, for example, transmission of endogenous patterns. While real-time feedback interventions driven by computer recognition of activity patterns are increasingly being applied to electrical stimulation (e.g., deep-brain stimulation) systems, most tools that have genetic precision and molecular specificity are still regulated en masse by rising gradients of sustained chemical drivers or by imposed patterns of optogenetic drive. Tools that permit direct experimental control of the efficacy and form of synaptic transmission between specific partners, will be a key step in providing the next wave of insight into network dynamics and function.
The most common strategy currently in use to modulate specific synaptic connections (Fig. 1a) involves light-activation of opsin-expressing presynaptic neurons using localized fiber optic stimulation near the postsynaptic target neurons (Fig. 1b 4 ). A high degree of presynaptic specificity can be achieved by this approach, using retrograde and Cre-dependent expression of optogenetic elements and localizing the light source, yet postsynaptic specificity will depend on the extent of presynaptic contact spread. A conceptually similar approach is achieved using chemogenetic methods to modulate synaptic transmission by targeting axon terminals (Fig. 1c). Chemogenetic neuromodulation can be restricted to a subset of target neurons, leaving other interconnected areas unaffected. For example, designer-receptorsactivated-exclusively-by-designer-drugs expressed in long-range projecting neurons in one cortical layer can be activated by ligand application to a confined area in another cortical layer 5 . These methods are limited by studying cell-cell communication between populations that are anatomically separate because 1 mm 3 of light (optogenetic) or exogenous drug (chemogenetic) will likely act on multiple cells. Further, most neural computations take place between highly interspersed cells in tightly packed spaces.
Here, we describe the Optical Synapse, an approach to control synaptic connectivity that utilizes presynaptically originating bioluminescence to activate optogenetic actuators expressed at postsynaptic sites (Fig. 1d). We have shown that bioluminescent optogenetics can operate within a cell, with a luciferase tethered to an opsin (a Luminopsin) [6][7][8][9] . Photon release from the luciferinluciferase interaction activates the associated opsin thereby achieving optogenetic modulation. Depending on the biophysical properties of the opsin, excitatory or inhibitory, bioluminescence may depolarize or hyperpolarize the neuron.
In our Interluminescent Optical Synapse, we used bioluminescent optogenetics to achieve synapse-specific and activitydependent circuit control, by expressing the luciferase presynaptically and its partner opsin postsynaptically. When presynaptic luciferase and postsynaptic opsin are present at the same synapse, bioluminescent optogenetic modulation is triggered when luciferin is provided, a requirement that allows experimental control of intersectional communication. The spatial requirements of Interluminescence in the current application restricts it to synapses that express the luciferase in presynaptic vesicles and opsins postsynaptically. The release of luciferase from the presynaptic terminal depends on presynaptic depolarization, similar to synaptic transmission via neurotransmitters or neuropeptides. Here we describe examples A expresses opsin (blue). LED light (light blue) drives all synaptic connections within its beam (driving B, but also C and D, and pattern imposed by experimenter).
A expresses DREADD (dark green). Local CNO injection (light green) enhances all synaptic connections within volume of injected drop (driving B, but also C and D).
A expresses luciferase (dark blue) and B opsin (blue). Activity (AP) of neuron A + CTZ creates light following synaptic release in all synapses, but only modulating targets expressing opsin (B only). Opsin expressing cells that are not contacted by the input partner of interest (A) also are not modulated (e.g., F). of the transmission of bioluminescence signals across synapses in culture and in vivo, and through a series of independent tests show that the postsynaptic output is mediated by optical coupling, and independent of classic neurotransmitter-mediated synaptic transmission. This intersectional technology can provide a novel class of cell-pairing and activity-specific control for testing the mechanisms underlying behavior.

Results
Interluminescence via an Optical Synapse. We used cultured cortical neurons to first establish if a luciferase genetically targeted to presynaptic vesicles could be released into the synaptic cleft and generate sufficient photon density to activate opsins genetically targeted to a postsynaptic cell. We term this form of bioluminescencemediated synaptic transmission Interluminescence (Fig. 2a). Critically, we needed to show that Interluminescence: (i) generates a measurable postsynaptic response; (ii) mediates different postsynaptic responses depending on the type of opsin, with cation permeable opsins triggering postsynaptic depolarization and excitation, and anion permeable opsins hyperpolarization and inhibition; (iii) occurs when luciferase is co-released with endogenous transmitters and peptides, but only in the presence of the luciferin; (iv) occurs independent of classic neurotransmission; and, (v) can co-exist with classic neurotransmission.
Interluminescence modulates spontaneous neural activity in culture. We targeted the blue light-emitting luciferase sbGLuc, a bright Gaussia luciferase variant 10 , to vesicles in cortical neurons using the vesicle targeting sequence of the human proopiomelanocortin pro-peptide (hPOMC1-26) 11,12 . The targeting construct also contained the reporter gene dTomato attached to sbGLuc via a P2A cleavage sequence (Fig. 2a). In addition to being a bright photon source, sbGLuc is favorable because it is also stable at the lower pH levels of synaptic vesicles 13,14 .
We selected high light sensitivity opsins for our initial experiments. We employed the excitatory step-function opsin ChR2(C128S) 15 and the inhibitory anion channel hGtACR2 16 as both exhibit high sensitivity to blue light relative to other opsins. In our initial studies, we separately nucleofected cortical neurons with hPOMC1-26-sbGLuc, ChR2(C128S), or hGtACR2 to ensure that luciferase and opsin were expressed in different cell populations (Fig. 2b). We then mixed the two populations of cells, one luciferase expressing and one opsin expressing, and plated the mixture on multielectrode arrays (MEAs) 17 . We used externally presented blue light to activate opsins directly and showed that this increased ( Fig. 2c middle panel; ChR2(C128S)) or decreased (Fig. 2c, lower panel; hGtACR2) the activity of the culture, as expected based on the type of opsin expressed. We then added the luciferin for Gaussia luciferases, Coelenterazine (CTZ), and observed increased (ChR2(C128S)) or decreased (hGtACR2) spontaneous activity consistent with the expressed opsin (Fig. 2c). By contrast, the addition of the vehicle alone had no consistent impact on ongoing neural activity.
We used direct LED stimulation of postsynaptic opsins as an internal control and compared the responses of cultures to LED (blue light), bioluminescence (CTZ), and control (vehicle) in three independent experiments (summarized in the ladder plots in Fig. 2d; Supplementary Data 1). Spontaneous activity levels varied across MEAs as expected (see also Supplementary Fig. 1). We measured spike rate (number of spikes/s) before and after treatment to compare each manipulation across different electrodes and MEAs. We observed a significant difference in spontaneous activity in cultures before and after stimulation by blue light LED and CTZ but no consistent differences in neural activity before and after vehicle (ChR2(C128S), blue light, n = 62,  . Our data also show that the activity of opsin-expressing postsynaptic neurons could be augmented, or inhibited depending on the nature of the postsynaptic opsin, following presynaptic activation in the presence of CTZ.
Interluminescence requires connections between pre-and postsynaptic neurons. To assess the properties of CTZdependent responses in more detail, and directly test if synaptic connectivity is required, we employed a co-culture system in which presynaptic and postsynaptic populations are seeded in separate compartments and then allowed to form synapses across a separating gap (Fig. 3a). We nucleofected presynaptic cortical neurons with luciferase, and postsynaptic hippocampal or striatal target neurons with ChR2(C128S).
We observed an increase in spontaneous activity and in spiking synchrony within the MEAs as the cultures matured ( Supplementary Fig. 2). The physical separation of pre-and postsynaptic neurons allowed us to identify the active population in different treatment conditions ( Supplementary Fig. 3). Direct electrical stimulation of presynaptic neurons increased spiking in ChR2(C128S) expressing postsynaptic neurons presumably through classic synaptic neurotransmission ( Supplementary Fig. 3c). Blue light LED stimulation (which directly activates the postsynaptic Opsin) or CTZ addition (which depends on ongoing spontaneous presynaptic activity) by contrast, increased spiking in ChR2(C128S) expressing postsynaptic neurons but without a change in the activity of cortical presynaptic neurons ( Supplementary Fig. 3d, e). We also showed that only MEA electrodes that responded to direct optogenetic stimulation (LED) responded to CTZ application; an independent demonstration of the specificity of CTZ-dependent responses. We next tested whether CTZ-dependent changes in postsynaptic neural activity were mediated by synaptic events triggered by presynaptic depolarization. We compared responses before and after severing the connections between the pre-and postsynaptic populations (Fig. 3b, upper versus lower panels). As expected, LED stimulation, which activates postsynaptic opsins directly, induced increased postsynaptic spiking equally in uncut and cut co-cultures. In contrast, direct electrical stimulation of presynaptic neurons and CTZ-induced bioluminescence failed to alter the excitability of postsynaptic neurons in cut co-cultures. This result is consistent with the hypothesis that CTZ modulation of postsynaptic activity requires synaptic connectivity. The results from several independent experiments of presynaptic (cortical) and postsynaptic (hippocampal, striatal) neural cultures are summarized ( Fig. 3c; blue light, un-cut n = 63, p < 0.0001, cut n = 58, p < 0.0001; electrical stimulation, un-cut n = 54, p < 0.0001, cut n = 32, p = 0.0965; CTZ, un-cut n = 51, p < 0.0001, cut n = 38, p = 0.7388; Wilcoxon matchedpairs signed rank test; Supplementary Data 2).
In summary, these data provide strong support for Interluminescence as a form of engineered synaptic transmission that achieves cell-cell communication via bioluminescence originating presynaptically activating opsins postsynaptically.
Interluminescence depends on presynaptic activity and occurs independent of classic synaptic neurotransmission. We found that addition of a cocktail of transmitter receptor blockers including: NBQX, a selective blocker of non-NMDA mediated synaptic transmission; D-AP5, a NMDA receptor antagonist; Gabazine and CGP55845, antagonists at GABA A and GABA B receptors, respectively; and Strychnine, a glycine receptor antagonist, silenced spontaneous activity of the entire neural culture on MEAs for several minutes ( Supplementary Fig. 4a, b(i), c; n = 27, P < 0.0001; Mann-Whitney test; Supplementary Data 3), but activation of postsynaptic opsin-expressing neurons by blue light LED confirmed that neurons responded to direct stimulation (Supplementary Fig. 4b(ii), c; n = 27, P < 0.0001; Mann-Whitney test; Supplementary Data 3). We employed neurotransmission blockade to test the requirements for bioluminescence-mediated synaptic transmission separate from endogenous transmitter mediated effects (Fig. 4). When CTZ was delivered with, or immediately after, addition of synaptic blockers (SB), postsynaptic activation through bioluminescence was robust, while vehicle addition had no effect ( These findings support the following model: In spontaneously active cultures luciferase is present in the synaptic cleft for a certain time period after its release. When CTZ is added immediately after acute synaptic block there is still sufficient luciferase to create photon density great enough to enable opsin activation. In contrast, within 20 s of synaptic block, luciferase levels in the synaptic cleft fall below that which could support opsin activation (silent phase). This interpretation is further supported by the observation that CTZ can induce spiking during the silent phase if it is added immediately following presynaptic electrical stimulation which triggers luciferase release into the synaptic cleft (Fig. 4d, e; e(iii): n = 10, electrical stimulation (after) v/s immediately added CTZ (after) p < 0.0001; Mann-Whitney test; Supplementary Data 4), while blockers (after) with electrical stimulation alone (after) or electrical stimulation (after) with immediate vehicle addition (after) had no effect (Fig. 4d, e; e(ii): n = 35, p = 0.9999; e(iv): n = 35, p = 0.8553; Mann-Whitney test; Supplementary Data 4). Increases in neural activity of opsinexpressing populations is observed in response to CTZ, but not to vehicle application immediately following electrical stimulation, nor when CTZ or electrical stimulation are applied by themselves ( Fig. 4e; yellow bars; immediate CTZ (after), n = 10, v/s immediate vehicle (after), n = 35; CTZ added later (after), n = 18; electrical stimulation (after), n = 35; p < 0.0001; Mann-Whitney test). These findings show that Interluminescence depends on presynaptic activity but that it is independent of the postsynaptic actions of synaptic neurotransmitters on neurotransmitter receptors.
Interluminescence depends on presynaptic vesicle release. To test whether presynaptic vesicle fusion is essential for Interluminescence, we used botulinum toxin (BoNT) to inhibit this process. BoNT, a neurotoxin that cleaves SNARE proteins, inhibits vesicular fusion and cargo release from presynaptic vesicles 18 (Fig. 5a). In BoNT-treated cultures, the level of luciferase in the media was reduced compared to untreated cultures, consistent with a decrease in vesicle release (Supplementary BoNT inhibits exocytosis of both small synaptic vesicles that contain small-molecule transmitters and large dense-core vesicles (LDCVs) that contain peptides. We used the sorting signal for the neuropeptide, POMC, to target luciferase to LDCVs using an hPOMC1-26-sbGLuc-eGFP fusion protein. We assessed the localization of sbGLuc-eGFP utilizing an antibody to dopamine β-hydroxylase which labels LDCVs 19 ( Supplementary Fig. 6). We observed partial colocalization of eGFP and anti-dopamine β-hydroxylase, confirming the presence of sbGLuc in dopamine β-hydroxylase expressing synaptic vesicles, but we also observed sbGLuc-eGFP signals consistent with a considerable fraction of luciferases in other vesicles.
Interluminescence-mediated activation of postsynaptic neurons requires opsins. We performed two experiments to test if bioluminescence-mediated activation of postsynaptic neurons occurs through opsins. First, we inactivated the step-function opsin ChR2(C128S) by exposure to longer wavelength light (590 nm). This protocol resulted in complete opsin inactivation (Fig. 6a schematic, Fig. 6b upper trace). To test if this inactivation also blocked Interluminescence, we applied synaptic blockers and CTZ to spontaneously spiking neurons, resulting in a neurotransmitter independent increase in spiking through bioluminescent activation of the opsin (Fig. 6b, lower trace). 590 nm light completely abolished spiking initiated by CTZ (Fig. 6b, lower trace) consistent with a need for recruitable opsins for Interluminescence ( Interluminescence in vivo: induction of cell-partner specific brain dynamics. An Interluminescence approach holds substantial distinct advantages for understanding behavior, as the complex processes underlying activities such as choice, memory and selective sensory processing inherently depend on cell-typespecific interactions between multiple brain areas. These interactions relay specific signals and create the dynamic states that facilitate or suppress specific channels of information. Such interareal processing is highly dependent on the specific type of neurons engaged in each area. A prominent example of this kind of cell-type-specific dynamic is gamma oscillations, rhythmic patterns of activity (~30-100 Hz) that are predictive of successful sensory processing [21][22][23] , and are believed to amplify the relay of sensory neural signals. Neocortical gamma depends on recruitment of fast-spiking, parvalbumin-positive (FS/PV) interneurons, either through endogenous or artificially applied glutamatergic drive, or by selective optogenetic activation [24][25][26] . Attentional gating of gamma in a given neocortical area is believed to be caused by excitatory intracortical 27,28 or thalamocortical [29][30][31] projections that recruit local FS/PV dynamics.
Given the potential utility of Interluminescence for in vivo studies, a crucial test is whether it can change network dynamics created by long-range, cell-type-specific communication. To test in vivo efficacy, we expressed the transmitting hPOMC1-26-sbGLuc in glutamatergic thalamic neurons (including ventral posterior medial and posterior medial) that target primary somatosensory neocortex (SI), including direct synaptic input to FS/PV 32,33 . The receiver excitatory opsins (ChR2(C128S/D156A)) were expressed under PV-Cre mediated control in SI FS/PV 24 ( Fig. 7a and Supplementary Fig. 7; this subset of mice will be referred to as Opsin (+)). In SI superficial and granular layers, PV is nearly exclusively FS-type 24 . In a control group, we expressed hPOMC1-26-sbGLuc in thalamus but not the excitatory opsin in neocortex (Opsin (−)).
In both Opsin (+) mice and Opsin (−) control mice CTZ presentation caused a rise in bioluminescent light production ( Fig. 7b; Supplementary Data 7). In Opsin (+) mice, we observed robust, broad-band gamma oscillation emergence that initiated with CTZ presentation (Fig. 7c; Supplementary Data 7). In contrast, similar changes were not observed in the Opsin (−) To systematically quantify these differences, in each mouse in the Interluminescence and control groups we calculated the overall increase in gamma power on each electrode compared to baseline for the period 0-1200 s after bioluminescent signal onset, and the probability of significant gamma events defined as ≥ 100 ms of increased gamma (see "Methods" for details). As shown in Fig. 7e In many cases, such as during the allocation of attention, increased local action potential firing is closely tied to increases in gamma-band activity 21 . Organization of FS/PV activity into a gamma pattern is believed to enhance signal relay by creating windows of opportunity for increased local firing, and increased firing due to rebound excitation 34 . Further, increased spiking associated with gamma increases may also reflect the disproportionate contribution of FS to multi-unit activity (MUA) measures, as FS/PV firing rates are typically an order of magnitude higher than those of nearby pyramidal neurons 35,36 . That said, FS/PV evoke powerful, soma-targeted inhibition, and robust recruitment of this cell class can create local suppression of spiking 32,35 , e.g., through optogenetic FS/PV recruitment at higher light intensities 24 .
As shown in Fig. 7f, during CTZ presentation, Opsin (+) mice showed MUA increases from baseline that were higher than, or equivalent to, the highest mean values of Opsin (−) mice. However, only one mouse showed a significant separation from the group. Accordingly, MUA differences were significant when analyzed at the electrode, but not the mouse, level (

Discussion
The present study shows that Interluminescence can provide a robust, synapse-selective, activity-dependent control of neural connectivity between specific pre-and postsynaptic partners. We show that presynaptic luciferase can activate postsynaptic opsins following presynaptic activity and only if luciferin is present. This chemogenetic element of the strategy provides an additional level of experimenter control. Interluminescence is modular, as light emission from the presynaptic luciferase can, in principle, activate any postsynaptic photoreceptor, including excitatory and inhibitory opsins and light-sensing GPCRs. Further, the approach is highly specific in that Interluminescence does not seem to have a volumetric effect.
Given its features, Interluminescence has the potential to provide a platform technology, in which the activity-dependence and postsynaptic impact of Optical Synapse recruitment can be selected. In this first instantiation described here, to maximize the coupling efficiency between luciferase and opsin, we used Gaussia luciferase, sbGLuc, which has high light emission, and the stepfunction opsin ChR2(C128S) and the anion channelrhodopsin 2 from Guillardia theta, hGtACR2 which have high photon sensitivity 6,8,9,20 . The precise photon density required to create the Interluminescence effects we observed is difficult to quantify in the abstract, as it depends on numerous factors including luciferase density, the impact of biologically specific variables in the cleft (e.g., pH sensitivity) and specific details of synaptic connectivity (e.g., synaptic distance, number of Interluminescent synapses expressing both components, location of these synapses on the postsynaptic cell, etc.). The luciferase-opsin combinations used here provide a baseline proof of concept that ample photon production was achieved in these in vitro and in vivo conditions, from a point of comparison for future luciferase-opsin pairings.
A prerequisite for Interluminescence is close proximity of the light emitter release presynaptically and light sensor located in the postsynaptic cell. In the instantiation of Interluminescence described here, we chose to express luciferase in synaptic vesicles, to concentrate light-producing enzymes to the presynaptic active zone and to ensure that presynaptic activity was required for luciferase release. We were able to take advantage of wellcharacterized pharmacological tools to manipulate vesicle exocytosis, including Botulinum Toxin (BoNT), thereby testing several assumptions including that Interluminescence should dependent on presynaptic vesicle fusion. BoNT inhibits both small synaptic vesicles and LDCVs 37 and immunohistochemistry revealed colocalization of luciferases in both dopamine β-hydroxylase-containing and non-dopamine β-hydroxylasecontaining vesicles 19 . Thus, although we used the POMC sorting signal to concentrate luciferase in peptide-containing LDCVs, our results suggest that luciferase was present in LDCVs as well as non-peptide-containing synaptic vesicles. Achieving a higher degree of specific targeting of luciferases to specific vesicles, either LDCVs or small synaptic vesicles could provide a way to establish functional connectivity via Interluminescence based on the stimulation frequency. For example, there is evidence that SVs are preferentially released in response to low-frequency stimuli compared to LDCVs which are preferentially released in response to higher frequency stimuli 38 . Synaptic vesicles are located in different subcellular domains of neurons including at presynaptic active zones, soma, dendrites, and axons. It is, therefore, possible and likely that luciferase is released at multiple sites following neuronal depolarization. Interluminescence likely reflects optical signaling at functional presynaptic synapses 39 because of the need for close proximity of luciferase and postsynaptic opsins across a shared synaptic cleft. Interluminescence outside of bona fide synapses is unlikely although this possibility could be explored in the future. Further, given that 20 s after release addition of luciferin does not elicit a postsynaptic response, it is most likely that luciferases diffused away from the synaptic space do not have a photon density high enough to activate opsins along the neuron. Thus, in contrast to neuropeptide transmission Interluminescence does not seem to have a volumetric effect.
The Interluminescence method described here is complementary to recently described orthogonal neuropeptide-receptor systems which regulate communication between genetically targeted preand postsynaptic partners. One is based on the insect peptide allatostatin and its receptor, both of which are inert in mammals 40 . The allatostatin receptor links via Gi/o-proteins to inhibit cyclic adenosine monophosphate (cAMP) and activate G-protein-coupled inward-rectifier potassium (GIRK) channels. Activity-dependent release of biologically active allatostatin from presynaptic neurons induces inhibition of allatostatin receptor-expressing subpopulations of postsynaptic neurons. The other system uses a Hydra derived presynaptically expressed neuropeptide and a matching postsynaptic cation channel that is opened by the peptide 41 . Upon activity-dependent presynaptic peptide release this heterologous synapse creates novel calcium fluxes postsynaptically and resulting in neural activation. Interluminescence, however, has some distinct advantages. First, Interluminescence is highly modular; luciferaseemitted light can be used to activate or inhibit partnering neurons depending on the opsin expressed, an advantage over an approach that requires separate systems for activation and inhibition. Second, Interluminescence utilizes opsins as universal current conductors, effecting direct changes in the membrane potential of the postsynaptic partner, an advantage over GPCR signaling pathways or Ca 2+ flux, both of which have the potential to engage a multitude of intracellular events. Third, transmission via synthetic chemical synapses is not highly restricted to presynaptic location, consistent with neuropeptide volume transmission. In contrast, luciferasedependent light emission decays over time and luciferases that diffuse beyond the synaptic cleft are unlikely to activate postsynaptic neurons. Fourth, synthetic chemical synapses are always on, whenever the presynaptic cell is active, and they are not under temporal control. By contrast, Interluminescence can be temporally gated by controlling luciferin availability, a feature advantageous for assessing the behavioral impact.
In summary, Interluminescence provides a unique technology for interrogating specific neural circuits with substantial temporal and spatial control. Interluminescence can boost or down regulate synaptic efficacy at specific synapses, it can be used to bias synaptic output, e.g. from inhibitory to excitatory and vice versa, and, in principle, it can establish new functional synaptic connections for example from silent to active. With rapid advances in the available palette of luciferases and opsins, this strategy can expand to meet a wide array of experimental needs.
In vitro Primary neurons. Primary neurons harvested from embryonic day 18 (E18) rat embryo cortex, hippocampus or striatum of both sexes were obtained from BrainBits, LLC, and processed according to the protocol provided by the company. Briefly, tissue was incubated for 10 min at 30°C in Hibernate E (minus calcium and B27 supplement; HEB, BrainBits) containing 2 mg/ml papain (BrainBits). Papain solution was removed, replaced by HEB medium, and tissue was triturated for about 1 min (90% tissue dispersal) using a 9″ sterile silanized glass Pasteur pipette (BrainBits), avoiding air bubbles. Undispersed pieces were allowed to settle for 1 min before the supernatant was transferred to a sterile 15 ml tube and spun at 1500 rpm for 10 min to collect the cell pellet. The pellet was resuspended in prewarmed and equilibrated NbActiv1 medium (BrainBits) and the cells were counted by Hemocytometer using Trypan blue stain.
Nucleofection. Nucleofection of E18 primary rat neurons was carried out per manufacturer's instructions (Amaxa Rat Neuron Nucleofector Kit # VPG-1003). Briefly, 1 × 10 6 primary neurons were collected and resuspended in 100 μl of Nucleofector Solution at room temperature. The cell suspension was combined with 1 μg plasmid DNA and transferred to the nucleofection cuvette. The Nucleofector 2b Device (LONZA # AAB-1001) was used for nucleofection with Nucleofector Program G-013.
MEA recordings. MEA2100-Lite-System (Multichannel Systems, Germany) was used for all MEA recordings. Consistently spiking neurons were used for recordings between DIVs 14-25 for the mixed and co-culture set-ups; only cultures showing spontaneous electrophysiological activity were used. All-trans retinal (R2500; Sigma-Aldrich, St. Louis, MO) was added to the culture medium to 1 μM final concentration before electrophysiological recordings. Prior to recording, all reagents were pre-warmed to 37°C. MEAs were transferred from the CO 2 incubator to the heated MEA2100 head stage maintained at 37°C, and the cultures were allowed to equilibrate for 5-10 min. The head stage was situated on a microscope stage (Zeiss Observer 1) with a fluorescent light source, allowing light stimulation of cultures at different wavelength through the objective. A micropipette was used to add reagents with the reagent drop gently touching the liquid surface, creating a time-locked artifact in the recordings. Recordings were carried out with a sample rate of 10,000 Hz. After recording, the media in the wells was replaced with fresh pre-equilibrated and pre-warmed NB-Plain media, and cultures were used for another round of recording the next day. MC Rack software was used for data acquisition. All MEA analysis was done offline with MC Rack software (Multichannel Systems; RRID: SCR_014955) and NeuroExplorer (RRID: SCR_001818). Spikes were counted when the extracellular recorded signal exceeded 9 standard deviations of the baseline noise. For assessing the effects of CTZ (10μm final concentration), only electrodes displaying the expected change in spiking activity with blue light from the fluorescent light source, i.e. opsin-expressing neurons, were evaluated. Pooled data were obtained from different electrodes (a) of the same culture, (b) from different cultures, and (c) over different DIVs.
Electrical stimulation. Electrical stimulation on MEA co-cultures was carried out using the integrated stimulus generator in the MEA head stage (MEA2100 Stimulator). The burst stimulation pattern was selected for a 100 μA current stimulus train, with the inter-pulse interval of 10 ms, and the pulses within this train were repeated 5 times.
Un-cut vs Cut experiment. Co-cultures were allowed to mature until there was synchronous firing activity across the co-culture. Effects of blue light, current stimulation, and CTZ were recorded from these synaptically connected un-cut cocultures. Thereafter, in the same co-cultures, inter-population connections were severed by running a piece of thin silicon like an eraser along the midline between the two populations. These cut co-cultures, which had lost the inter-population synchronicity, were then subjected to the same treatments (blue light, current stimulation and CTZ).
Synaptic blockers. The cocktail of synaptic blockers (SB) (final concentrations indicated) included NBQX (10 μM), D-AP5 (50 μM), Gabazine (100 μM), CGP55845 (100 μM) and Strychnine (1 μM). Aliquots were stored at −80°C and each time thawed freshly right before the start of the MEA recording. The SB cocktail was incubated at 37°C before being added to the MEA and was added gently as 10 μl drop to the neuronal media in the MEA well. For recordings involving a mixture of SB cocktail with either CTZ or vehicle, CTZ or the vehicle stocks were freshly diluted with the SB cocktail to attain the final CTZ concentration of 10 μM or equivalent in case of the vehicle.
Botulinum neurotoxin (BoNT). BoNT/A1 was used as a blocker for the vesicular release of the presynaptic luciferase. BoNT was used at 30 ng/ml for 48 h before recording experiments.
Confocal Imaging. For confocal microscopy nucleofected cortical neurons were grown on Poly-D-Lysine coated coverslips (Neuvitro GG-12-PDL) in 24-well dishes until DIV 21. Neurons were fixed by completely removing the media from each well and then adding 500 µL of 4% paraformaldehyde and incubating for 15 mins at RT, followed by 3 washes for 5 min each in PBS. Neurons were permeabilized by incubating in 0.1% Triton X-100 and again were washed 3 times for 5 min each in PBS. Neurons were blocked with 1% Bovine Serum Albumin (BSA) in PBST (PBS + 0.1% Tween 20) for 1 hr, incubated for 12 h at 4°C with a rabbit polyclonal anti-Dopamine β Hydroxylase (DβH) antibody (Millipore Sigma, AB 1585, diluted 1:2000 in 1% BSA in PBST), then washed 3 times, for 5 min each time, with PBST. Neurons were then incubated for 1 hr at RT with Donkey anti-Rabbit IgG H&L (Alexa Fluor 594; ab150076; diluted 1:500 in 1% BSA in PBST) and washed 3 times for 5 min each in PBST. Cells were mounted in antifade mounting media (Vectashield Hardset, H-1500-10) containing DAPI and imaged with a Nikon A1 confocal laser scanning inverted microscope using a Nikon Plan Apo VC 60x/1.40 Oil DIC N2 objective (1024 × 1024 µm). To image sbGLuc-eGFP the optical sections were scanned with the 561 nm laser line at 60% intensity. To detect DβH + dense core vesicles with Alexa Fluor 594 the 561 nm laser line at 2% intensity was utilized. Detection was done with a 450/50 filter cube for eGFP and 595/50 for Alexa Fluor 594. The raw images were exported as TIF files and analyzed with ImageJ (Rasband, W.S., ImageJ, U.S. National Institutes of Health, Bethesda, Maryland, USA, http://imagej.nih.gov/ij/).

Statistics and reproducibility.
All analyses were performed with Prism software (GraphPad 8.2.1; San Diego, CA), which provides the evaluation of the suitability of the test for the specific data set. MEA data were collected from multiple recordings within each experiment and from multiple experiments. For randomization, the plates were switched for different treatment conditions, e.g., the plate used for CTZ treatment on one day was used for vehicle treatment the next day and vice versa. Positive control blue light treatment using the fluorescent light source was done for each recording along with other treatments as the basis for selecting electrodes for analysis of opsin-expressing postsynaptic neurons. All the treatments (e.g. CTZ vs vehicle) were carried out on similar DIVs (13-21 for mixed cultures; 21-26 for co-cultures) to control for age and synaptic connectivity-related variations within neuronal cultures on MEAs. For each MEA data ladder plot, n = number of electrodes were assessed. The differences in number of spikes before and after treatment were assessed for significance. Due to non-normal distribution of data, non-parametric paired Student's t tests (two-tails) were used. To evaluate the within-group differences, Wilcoxon matched-pairs signed-rank test was used, and to evaluate across-groups differences, Mann-Whitney U tests were used with significance set at p < 0.05 (ns, not significant; *p < 0.05; **p < 0.002; ***p < 0.0002; ****p < 0.0001) using 95% confidence level. Throughout the paper, the medians are highlighted for each ladder plot in the figures. The time analyzed for the number of spikes before and after treatment was 5 s for all ladder plots, removing the time pertaining to artifacts due to the addition of reagents (indicated by the white gap area in the representative recording traces as noted in the figure legends). n and p values and the type of statistical test used are described for each ladder plot in the figure legends and results.

In vivo
Animals. Six PV-Cre mice (all male; JAX stock #008069) aged 9 to 19 weeks (M = 16.10, SD = 3.72) were used. Three mice were injected with the luciferase virus (AAV2/9-hSyn-hPOMC1-26-sbGLuc-P2A-dTomato) in somatosensory thalamus along with injections of a Cre-dependent excitatory step-function opsin in SI (Opsin (+) animals). As a control, three additional mice were also injected with the luciferase virus in somatosensory thalamus, but no opsin was introduced (Opsin (−) animals). The Opsin (−) animals thus should produce light in SI due to thalamocortical projections to SI, but no optogenetic effect ought to occur because of the absence of a postsynaptic opsin. Imaging data are unavailable for one of the Opsin (+) animals due to a software malfunction. Mice were housed in a vivarium on a reversed light-dark cycle and had free access to food and water. All procedures were conducted in accordance with all relevant ethical regulations for animal testing and research and under a study protocol approved by the Animal Care and Use Committee of Brown University.
Surgeries and course of experiment. Approximately three weeks prior to the day of the experiment, each animal was anesthetized (~1% isoflurane), fitted with a steel headpost, and injected with viral constructs via burr holes made with a dental drill. Animals received an injection of 400 nl of luciferase virus (AAV2/9-hSyn-hPOMC1-26-sbGLuc-P2A-dTomato) into the somatosensory thalamus (−1.75 A/ P ± 0.05, M/L 1.575 ± 0.175, D/V −3.4 relative to Bregma). This injection strategy targeted somatosensory thalamus broadly, likely infecting neurons in both the ventral posterior medial and the posterior medial nuclei. All viral injections were performed through a glass pipette fitted in a motorized injector (Stoelting Quintessential Stereotaxic Injector, QSI). The Opsin (+) animals also received additional viral injections of 200 nl of the excitatory step-function opsin (pAAV-Ef1a-DIO hChR2(C128S/D156A)-EYFP; a gift from Karl Deisseroth; Addgene viral prep # 35503-AAV1; RRID:Addgene_35503) in three locations of SI equidistantly spaced around a central SI point (−1.25 A/P, 3.25 M/L) at a depth of 350 µm. All viral constructs were delivered at a rate 50 nl/min. After 2-3 weeks of recovery, and to allow for viral expression of the constructs, experiments were conducted under isoflurane at~1% (0.5-2%). A dental drill was used to make a 3 mm diameter circular craniotomy centered over SI (−1.25 A/P and 3.25 M/L relative to Bregma). The exposed brain remained covered in saline throughout the experiment. The animal was moved to a light tight and electrically shielded box and continued to receive anesthesia. A 32-channel probe was inserted into the cortex perpendicularly to the cortical surface at a rate of~10 µm/s using a motorized micromanipulator (Siskiyou MD7700) to a depth of 795 µm or until the highest contact on the probe disappeared from view into the cortical tissue as viewed from a stereoscope. The probe was then allowed to rest in this position for 30 min before starting the experiment. After baseline recordings of a minimum of 3 minutes, the luciferin CTZ was introduced and recordings continued for a minimum of 20 minutes. At the conclusion of the experiment, mice were euthanized with isoflurane and perfused transcardially with 4% paraformaldehyde (PFA). The brain was removed and post fixed in 4% PFA at 4°C for approximately 48 h after perfusion. The brain was then placed in 30% sucrose at 4°C for a minimum of 36 h before slicing. Brains were then sectioned at 50 µm on a cryostat (Leica CM30505) and mounted on glass slides. Fluorescent tags in the sectioned brains were imaged on a Zeiss LSM 800 confocal microscope to verify correct viral targeting.
Luciferin delivery. Water soluble coelenterazine (Nanolight #3031) was dissolved in sterile water (1 µg/ml) to yield a final concentration of 2.36 mM. The solution was loaded into a 250 µl glass syringe (Hamilton #80701) fitted with a~1 cm length of 18-gauge plastic tubing. The Hamilton syringe and tubing were placed in a motorized injector (Stoelting Quintessential Stereotaxic Injector, QSI). The tip of the plastic tubing was lowered into the pool of saline over the craniotomy using a micromanipulator until it touched the surface of the skull. The tip of the tubing was further adjusted so that it rested at a distance of~3 mm from the opening edge of the craniotomy. The luciferin CTZ was delivered by infusing 50 µl of the solution into the saline over the open craniotomy at a rate of 25 µl/min. Imaging and electrophysiological recordings. Electrophysiological data were acquired using an Open Ephys acquisition board (http://www.open-ephys.org/) connected via an SPI interface cable (Intan) to a 32-channel head stage (Intan). A 32-channel laminar probe (Neuronexus, A1x32-Poly2-5mm-50s-177) was connected to the head stage. The iridium electrode contacts on the probe covered a linear length of 790 µm and were arranged into two columns of 16 contacts spaced 50 µm apart. The data were acquired using the Open Ephys GUI software at a sampling rate of 30 kHz and referenced to a supra-dural silver wire inserted over the right occipital cortex. Imaging data were acquired using an Andor iXon Ultra 888 EMCCD camera attached to a Navitar Zoom 6000 lens system. The data were acquired using Andor Solis data acquisition software (Andor Solis 64 bit, v4.31). The field of view was centered over the craniotomy and adjusted to encompass the full diameter of the craniotomy. Images (512 × 512 pixels,~6 µm 2 /pixel) were acquired continuously at an exposure length of 1 s and an electron multiplication gain of 300. The data were acquired in units corresponding to the number of electrons recorded by a given pixel. A TTL pulse synchronized the recording of the imaging and electrophysiological data.
Electrophysiology analysis. Offline analyses of both electrophysiological and imaging data were performed in MATLAB R2020a (The Mathworks Inc.). The electrophysiological data were down sampled to 10 kHz. For each recording, electrode contacts with RMS values more than three times the interquartile range above the 3rd quartile or three times the interquartile range less than the 1st quartile of all 32 electrode contacts were marked as errant and removed from further analyses. Across all animals a total of seven electrode contacts were marked bad, so all reported electrophysiological data are from the remaining 25 contacts. These electrode contacts were then re-referenced to the common average reference 43 .
For the time-frequency analysis of the local field potential, the data were further down sampled to 1 kHz and high-pass filtered with a cutoff at 1 Hz (3rd order Butterworth). Spectral analysis of the time series of each electrode contact was performed using a sliding multitapered fast Fourier transform using the Chronux software package for MATLAB (version 2.11, http://chronux.org) 44 . The timebandwidth product for the multitaper analysis was set to 3 and 5 tapers were used. Sliding windows of 10 s in steps of 1 s were employed to analyze the spectrotemporal evolution of the time series, and each 10 s window was zero padded to a total length of 2 14 = 16384 samples. Changes in spectral power relative to baseline for each electrode contact, time window, and frequency band were represented in decibel scale as follows: where B is the average power across time in the baseline period, defined here as the 180 s period prior to the onset of bioluminescence, for a given electrode contact and frequency band, and A is the power in a given electrode contact, frequency bin and time bin. In addition to estimating the overall power change in the gamma band we also assessed whether high-power, but brief events in the gamma band may have increased with the introduction of the CTZ. Such events may be less detectable by the multitaper analysis due to the long time windows employed. To detect such short lived gamma events we bandpass filtered the data between 30 and 100 Hz (3rd order Butterworth), Hilbert transformed the data to acquire the analytic signal and took the absolute value to acquire an estimate of the instantaneous amplitude envelope in the gamma band. Next gamma events in the post CTZ period were first defined as any data point that exceeded the 99th percent jackknifed confidence interval of the baseline mean amplitude. An event was then required to exceed this threshold for at least 100 ms (i.e., at least 3 cycles at 30 Hz). Using these criteria, events were summed across the 1200 s post CTZ period for each electrode contact.
To isolate MUA, a bandpass filter (passband: 300 Hz to 3000 Hz, 3rd order Butterworth) was applied to the data. Spikes were defined as data points less than −3 times the standard deviation, where the standard deviation was estimated as the median divided by 0.6745 18 . Spikes were then binned in increments of 1 s to yield a MUA time series for each electrode contact. Each time series was then converted to percent change from baseline.
Bioluminescence imaging analysis. For all images, a 3 × 3 pixel median filter was applied to reduce shot noise. For each animal, a circular region with a diameter of 50 pixels was placed in the region directly adjacent to the electrode shank and in front of the surface with the exposed electrode contacts. The mean of these pixels was computed for each image to yield a time series of bioluminescence. Since CTZ was infused into the saline over the craniotomy there was some variability in the onset of bioluminescence across animals. We were specifically interested in the relationship of bioluminescence to changes in gamma-band activity and MUA, so we aligned all data to the onset of bioluminescent signal in the imaging data. We quantified the onset of bioluminescence as the peak of the discreet derivative of the bioluminescent signal. This method worked well since the bioluminescent signals in these experiments were monotonically increasing with a rapid onset.
Statistical analysis. As an initial descriptive statistic we calculated the number of electrode contacts that exceeded baseline for each of the three dependent measures (gamma power (dB) relative to baseline, number of gamma events and MUA percent change from baseline) in the Opsin (+) and Opsin (−) groups. A given electrode contact was considered to have exceeded baseline if its mean value in the period after bioluminescence onset was above the bootstrapped 95th confidence interval of the mean of the baseline period.
For each of the three dependent measures we initially acquired threedimensional matrices with dimensions electrode contacts, time bins, and animals. In the case of gamma power, the frequency dimension of the spectrograms was collapsed by averaging over the 30-100 Hz frequency bins. To test for significant changes in these measures we took a multi-level approach. First, for each of the dependent measures we pooled all data points across electrode contacts, time bins and animals, keeping the Opsin (+) and Opsin (−) groups separate. These pooled data sets were then submitted to a two-sample Kolmogorov-Smirnov (KS) test to test broadly for differences in the distributions of the Opsin groups. If a given KS test indicated a significant difference between the groups we then averaged across the time dimension and submitted the electrode contacts in each Opsin group to a Mann-Whitney U test. In the case of the gamma measures these tests were performed as one-tailed tests as we hypothesized a priori that upregulation of PV cells in SI via Interluminescence should increase gamma-band activity. The test of MUA differences was performed as a two-tailed test as we had no a priori hypothesis about the directionality of these effects. The KS tests coupled with inspection of the CDFs of the two groups distributions suggested that the Opsin (+) group was positively skewed relative to the Opsin (−) group. Therefore, to sensitively test for group differences at the animal level we computed the 85 th percentile value for each of the measures from the array of electrode contacts for each animal. These values were then submitted to Mann-Whitney U tests to test for significant differences between the Opsin group at the animal level. Again, gamma-band activity measures were submitted to one-tailed tests, while the MUA data were submitted to a two-tailed test.