Optogenetic Determination of Dynamic and Cell-Type-Specific Inhibitory Reversal Potentials

Abstract

The reversal potential refers to the membrane potential at which the net current flow through a channel reverses direction. The reversal potential is determined by transmembrane ion gradients and, in turn, determines how the channel's activity will affect the membrane potential. Traditional investigation into the reversal potential of inhibitory ligand-gated ion channels (E_Inh) has relied upon the activation of endogenous receptors, such as the GABA-A receptor (GABA_AR). There are, however, challenges associated with activating endogenous receptors, including agonist delivery, isolating channel responses, and the effects of receptor saturation and desensitization. Here, we demonstrate the utility of using a light-gated anion channel, stGtACR2, to probe E_Inh in the rodent brain. Using mice of both sexes, we demonstrate that the properties of this optically activated channel make it a suitable proxy for studying GABA_AR receptor-mediated inhibition. We validate this agonist-independent optogenetic strategy in vitro and in vivo and further show how it can accurately capture differences in E_Inh dynamics following manipulations of endogenous ion fluxes. This allows us to explore distinct resting E_Inh differences across genetically defined neuronal subpopulations. Using this approach to challenge ion homeostasis mechanisms in neurons, we uncover cell-specific E_Inh dynamics that are supported by the differential expression of endogenous ion handling mechanisms. Our findings therefore establish an effective optical strategy for revealing novel aspects of inhibitory reversal potentials and thereby expand the repertoire of optogenetics.

Significance Statement

The strength of synaptic inhibition in the brain is determined, in part, by the reversal potential of the ionic currents that flow through inhibitory ligand-gated ion channels (E_Inh). Estimates of E_Inh have traditionally used agonists to activate receptors on the cell surface, but this has limitations. Our study presents an optogenetic strategy for performing agonist-independent measurements of E_Inh in the brain. We demonstrate the effectiveness of the approach in vitro, in vivo, and across different neuronal subtypes. Its excellent temporal control allows for measurements of E_Inh dynamics, which reveal differences between genetically defined neuronal subpopulations. This expands the application of optogenetics and affords new opportunities to study synaptic inhibition.

Introduction

The reversal potential of current passing through inhibitory ligand-gated ion channels (E_Inh) affects the strength and nature of fast synaptic inhibition in the nervous system and is determined by transmembrane gradients for the ions that permeate these channels. In the brain, GABA-A receptors (GABA_ARs) are the principal ligand-gated ion channels that mediate fast synaptic inhibition, and these channels are predominantly permeable to chloride (Cl⁻) ions and, to a lesser extent, bicarbonate (HCO₃⁻) ions (Takeuchi and Takeuchi, 1967; Bormann et al., 1987; Kaila and Voipio, 1987). E_Inh is therefore primarily determined by the equilibrium potential for Cl⁻ (E_Cl), which results from the dynamic interplay between Cl⁻ extrusion and intrusion processes, including effluxes mediated by a potassium chloride cotransporter, KCC2 (Thompson et al., 1988; Rivera et al., 1999; Düsterwald et al., 2018) and influxes such as those mediated by GABA_ARs themselves (Doyon et al., 2011; Yelhekar et al., 2017). E_Inh is not a fixed parameter, but can vary between cells and over a range of timescales within the same cell, as a result of differences in transmembrane Cl⁻ gradients caused by relative changes in Cl⁻ influx and efflux (Staley and Proctor, 1999; Doyon et al., 2016b; Currin et al., 2020).

Estimates of E_Inh have traditionally been made by activating endogenous receptors on the cell surface and measuring current flow at different membrane potentials. This involves exogenous application of an agonist (focal delivery or uncaging of GABA; Alger and Nicoll, 1979; Ilie et al., 2012; Kirmse et al., 2015) or stimulation techniques that elicit the release of a relevant endogenous agonist (electrical or optical activation of presynaptic GABA-releasing axons; Staley and Mody, 1992; Kätzel et al., 2011; Zhao et al., 2011). There are, however, limitations associated with inferring E_Inh from endogenous receptors. Firstly, it can be difficult to be sure which receptor populations are being activated, which is problematic as E_Inh is thought to vary between subcellular compartments (Szabadics et al., 2006; Földy et al., 2010; Ellender et al., 2014). It can be challenging to deliver the agonist to the cell of interest in intact preparations, it is often necessary to use pharmacology to isolate the receptor's response, and it is difficult to determine the receptor's own contribution to E_Inh (Karlsson et al., 2011; Yelhekar et al., 2016). Furthermore, additional processes influence attempts to measure dynamic changes to E_Inh. Agonist application can saturate receptors, desensitization processes can alter receptor availability (Frosch et al., 1992; Jones and Westbrook, 1995), and, when stimulating axonal release, presynaptic mechanisms can alter how the agonist is delivered (Kullmann et al., 2005).

Optogenetics has provided neuroscience with an invaluable toolkit for modulating neuronal activity. Since these tools utilize light-activated ion fluxes, they also afford an untapped opportunity to examine transmembrane ion gradients. While light-activated anion pumps have been shown to alter E_Inh (Raimondo et al., 2012), a more appropriate strategy for measuring E_Inh would be to use a light-gated channel that passively reports transmembrane ion gradients, such as an anion channelrhodopsin (ACR; Berndt et al., 2014; Wietek et al., 2014). Within this class of opsins, Guillardia theta anion-conducting channelrhodopsin 2 (GtACR2) has excellent performance characteristics due to its large photocurrents and ion selectivity (Govorunova et al., 2015; Wiegert et al., 2017).

Here, we establish GtACR2's use as an agonist-independent strategy for probing E_Inh in the mouse brain. This allows us to quantify the contribution of endogenous ion fluxes to E_Inh and is leveraged to distinguish E_Inh in different neuronal subpopulations. By using this approach to challenge ion homeostasis, we uncover cell-specific E_Inh dynamics that are supported by the differential expression of endogenous handling mechanisms. This generates new opportunities for investigating E_Inh using light, in a manner that is compatible with the targeting of defined cell populations and subcellular compartments.

Materials and Methods

Intracortical viral injections

To selectively express the soma-targeted light-gated stGtACR2 anion channel in excitatory or inhibitory cortical neuronal subpopulations, the Cre-dependent construct pAAV_hSyn1-SIO-stGtACR2-FusionRed packaged in an adenoassociated virus 1 (AAV1) was injected into the primary somatosensory cortex (S1) of mice homozygous for either CaMKIIα-Cre (Tg(Camk2a-cre)T29-1Stl/J; The Jackson Laboratory, RRID:IMSR_JAX:018966) or PV-Cre (B6;129P2-Pvalb^tm1(cre)Arbr/J; The Jackson Laboratory, RRID:IMSR_JAX:008069). The pAAV_hSyn1-SIO-stGtACR2-FusionRed was a gift from Ofer Yizhar (Addgene plasmid #105677; http://n2t.net/addgene:105677; RRID:Addgene_105677). Mice (4 weeks, P28) were anesthetized with isoflurane (Zoetis) and placed in a stereotaxic frame (Kopf). Analgesic management included intravenous meloxicam (5 mg/kg) and buprenorphine (0.1 mg/kg) as well as intradermal Marcain (2 mg/kg) injected into the scalp. Each cornea was protected with ointment (Viscotears), and the animal's temperature was monitored throughout the procedure. The scalp was shaved (Wahl) and cleaned with a chlorhexidine-based solution (Hibiscrub). A midline incision was made that extended from the interpupillary line to the interaural line, and careful dissection was performed to expose the underlying cranium. The craniotomy (diameter <1 mm) was located over the left S1 (coordinates in relation to the bregma: 3 mm lateral, 1.2 mm posterior) and performed with a dental drill (Foredom). A beveled glass micropipette (Blaubrand IntraMARK) preloaded with undiluted virus was inserted 300 μm below the surface of the cortex. A total volume of 500 nl was injected at a rate of 33 nl/min. After every 100 nl, the micropipette was raised in 50 μm increments. Following completion of the injection, the micropipette was left in situ for an additional 5 min before it was removed. The scalp was sutured (6-0 Vicryl, Ethicon), and the animal was recovered before being returned to its home cage. Following injections, mice were maintained for at least 2 weeks.

In utero electroporation (IUE)

Surgery was performed upon timed pregnant female C57BL/6 wild-type mice, with the day of plugging defined as embryonic Day 0.5 (E0.5). IUE was performed at E15.5 to maximize targeting of cortical pyramidal neurons in L2/3 (Tabata and Nakajima, 2001; Ellender et al., 2019). On the day of the procedure, the pregnant dam was anesthetized with isoflurane (Zoetis) and given subcutaneous analgesia (5 mg/kg meloxicam and 0.1 mg/kg buprenorphine). A midline laparotomy was performed, and the uterine horns were exposed. DNA plasmids were loaded into a micropipette made from borosilicate glass capillaries (Harvard Apparatus). The DNA plasmids consisted of Cre recombinase under the chicken β-actin (CAG) promoter, which was a gift from Connie Cepko (Addgene plasmid #13775; http://n2t.net/addgene:13775; RRID:Addgene_13775); Cre-dependent stGtACR2-FusionRed under the hSyn1 promoter, which was a gift from Ofer Yizhar (Addgene plasmid #105677; http://n2t.net/addgene:105677; RRID:Addgene_105677); and the cytosolic reporter tandem tomato (tdTomato) under the CAG promoter, which was a gift from Angelique Bordey (Addgene plasmid #83029; http://n2t.net/addgene:83029; RRID:Addgene_83029). The plasmids were combined in the following amounts: 7 μl of CAG-Cre (3 μg/μl), 7 μl of stGtACR2 (2.6 μg/μl), and 5 μl of CAG-tdTomato (3.7 μg/μl). One microliter of 0.03% fast green dye (Sigma-Aldrich) was added to make a total volume of 20 μl. For each embryo, 1.5 μl of the plasmid-containing mixture was injected intraventricularly. The anode of a 5 mm platinum Tweezertrode (BTX) was then positioned over the dorsal telencephalon external to the uterine muscle, and five pulses of 36 V (each pulse 50 ms separated by a 950 ms interval) were delivered using a pulse generator (BTX). After all embryos had been electroporated, the uterine horn was returned to the abdominal cavity, which was filled with warm sterile saline solution. The abdominal wall was closed with Vicryl sutures (Ethicon), and the skin was closed with Prolene sutures (Ethicon). Dams were closely monitored during the recovery period. Pups were screened for successful electroporation on the third postnatal day (P3), by illuminating the head with a custom-built light-emitting diode (LED, 525 nm, 30 W) to detect tdTomato fluorescence. The pups were weaned at 3 weeks of age and maintained until 6–8 weeks for experiments.

Immunohistochemistry

Mouse brains were fixed via transcardial perfusion of phosphate-buffered solution (PBS, 0.1 M, Sigma-Aldrich) and 4% paraformaldehyde solution (PFA, Sigma). Brains were stored for 24 h at 4°C in 4% PFA and then washed and stored in PBS containing 0.05% sodium azide (Merck). Within a week of perfusion, brains were washed in PBS and mounted onto a microtome (HM650V, Thermo Fisher Scientific) before being sectioned into 100-μm-thick coronal sections while being bathed in PBS. Thereafter, sections were washed with fresh PBS before being washed with PBS containing 0.3% Triton X-100 (Sigma-Aldrich; 0.3% PBST). Sections were then blocked in a solution containing 20% normal goat serum (NGS) in 0.3% PBST for 2 h. The sections were washed with 0.3% PBST after blocking and then incubated at 4°C with the primary antibody solution containing 10% NGS, 0.3% PBST plus the relevant antibody for Cux1 (Santa Cruz, #SC-13024, RRID:AB_2261231) or PV (Synaptic Systems, #195004, RRID:AB_2156476). After 24 h, sections were again washed in 0.3% PBST before incubation with the secondary antibody at room temperature (∼25°C) for 2 h in a solution containing 5% NGS, 0.3% PBST and a fluorophore-linked secondary antibody specific for the relevant primary antibody. These included an anti-rabbit 635 for Cux1 labeling (Thermo Fisher Scientific, #A-31577, RRID:AB_2536187) and an anti-guinea pig 488 for PV labeling (Thermo Fisher Scientific, #A-11073, RRID:AB_2534117). Sections were washed in fresh PBS and mounted on slides with VectaShield (Vectorlabs). Confocal imaging was performed using a LSM 880 microscope (Zeiss) equipped with 488, 561 and 633 nm laser lines. All images were captured using a 20× water-immersion objective (W Plan-Apochromat NA 1.0) with the ZEN software (Zeiss). Image processing was performed in ImageJ software (NIH).

Acute slice preparation

In vitro electrophysiological experiments were performed in acute brain slices prepared from S1 of juvenile/adult mice aged 6–8 weeks, 2–3 weeks after intracerebral viral injection. Mice were first anesthetized with isoflurane (Zoetis) and decapitated. The brain was then sliced into 350 μm coronal sections using a microtome (HM650V, Thermo Fisher Scientific) in carbonated high-sucrose solution maintained at 4°C and consisting of the following (in mM): 185 mM sucrose, 1.2 NaH₂PO₄, 25 NaHCO₃, 2.5 KCl, 25 glucose, 2 CaCl₂ and 2 MgCl₂ (all Merck). The pH was adjusted to 7.3–7.4 (S20, Mettler Toledo) and the osmolarity to 300–310 mOsm (5520 Vapro, Wescor). Brain slices were recovered in an incubation chamber maintained at 37°C and containing artificial cerebrospinal fluid (aCSF) consisting of the following (in mM): 120 NaCl, 3 KCl, 1.2 NaH₂PO₄, 23 NaHCO₃, 11 d-glucose, 2 CaCl₂, and 2 MgCl₂ (pH, 7.3–7.4; mOsm, 300–310 mOsM). Slices were recovered for at least 60 min before starting electrophysiological recordings. When required, slices were transferred to a recording chamber that was continuously superfused with the bicarbonate-buffered aCSF and gassed with 95% O₂/5% CO₂ (32°C, pH 7.3–7.4, and perfusion speed of 2 ml/min).

In vitro electrophysiological recordings

All electrophysiological recordings were performed using an Axopatch 700 A amplifier with data acquired using the Clampex software (Molecular Devices). A HumBug noise eliminator (Digitimer) was used to remove 50 Hz noise. For in vitro whole-cell recordings, micropipettes (2–5 MΩ) with a short shaft were prepared from borosilicate glass capillaries (Harvard Apparatus) using a horizontal puller (Sutter) and filled with a K⁺ gluconate-based internal solution, which contained the following (in mM): 126 K-Glu, 4 KCl, 4 Na₂ATP, 0.3 NaGTP, 10 Na₂-phosphocreatine, 10 HEPES, and 0.05 EGTA. For experiments in which intracellular Cl⁻ was varied, this was achieved by adjusting the amount of K-Glu and KCl. For the 20 mM Cl⁻ internal solution, 110 mM K-Glu and 20 mM KCl were used. For the 70 mM Cl⁻ internal solution, 60 mM K-Glu and 70 KCl were used. All other reagents were maintained the same. For all internal solutions, pH was adjusted to 7.3 using K⁺ hydroxide (Mettler Toledo), and osmolarity was adjusted to 287–293 mOsm (Wescor). Pipettes were back-filled with the internal solution and mounted on a micromanipulator (Sutter), positive pressure was applied, and the pipette tip was lowered under visual guidance onto the surface of a neuron's soma within the brain slice. Neurons were visualized using dot contrast and a 60× objective (LUMPlanFI/IR, NA 0.90, Olympus). To perform perforated patch-clamp recordings, the internal pipette solution was prepared immediately prior to recording by combining a high Cl⁻ (150 mM) solution (in mM: 141 KCl, 9 NaCl, 10 HEPES) heated to 37°C with a stock solution of gramicidin A (4 mg/ml—dissolved in dimethyl sulfoxide, DMSO, Merck) to achieve a final concentration of 80 μg/ml gramicidin (Kyrozis and Reichling, 1995). The solution was then vortexed (40 s) and sonicated (20 s), before the pipette was back-filled with the gramicidin solution. Once the gigaseal had formed, perforation was monitored by observing changes in series resistance. Recording protocols were started once the series resistance had stabilized at <50 MΩ. Rupture or breakthrough of the perforation into whole-cell configuration was detected by sudden and persistent depolarization of the measured E_GABAAR, consistent with dialysis of the neuron with the high Cl⁻ pipette solution. stGtACR2 was activated by delivering blue pulses of light via a 1,000 μm optic fiber connected to a 473 nm laser (MBL-FN-473-150mW, CNI Laser). The tip of the fiber was positioned at an image plane in the microscope, in the center of the optical axis, and directed into the 60× objective lens via a dichroic mirror. This resulted in a spot of light at the brain slice whose diameter was 65 μm, assuming zero tissue scattering. Laser intensity was controlled via a potentiometer on the laser power controller, and the timing of laser pulses was controlled via TTL pulses delivered from an analog-to-digital board (Molecular Devices), which were synchronized with the electrophysiology software. In a subset of experiments, focal GABA (100 μM, Tocris Bioscience) was applied to the soma of the recorded neuron via a micropipette attached to a picospritzer (5 psi, 10 ms, General Valve). To isolate the GABA_AR response, GABA_BRs were blocked by adding a selective antagonist, CGP 55845 (Cunningham and Enna, 1996), at a final concentration of 10 μM in the circulating aCSF. In a subset of experiments, KCC2 was blocked by adding a selective antagonist, VU0463271 (Delpire et al., 2012), at a final concentration of 10 μM in the circulating aCSF.

In vivo two-photon-guided patch-clamp recordings

In vivo-targeted patch-clamp recordings were performed based on previously published protocols (Margrie et al., 2002; Wang et al., 2016; Jouhanneau and Poulet, 2019). Mice aged 6–8 weeks that had undergone IUE were anesthetized with an intraperitoneal injection of 25% urethane (1 g/kg in sterile PBS). To counteract adverse events caused by urethane, a bolus of the anticholinergic agent, glycopyrronium bromide (0.01 mg/kg), was administered subcutaneously. Local anesthetic (Marcain 2 mg/kg) was applied intradermally in the scalp and topically in the ears prior to mounting the mouse into the head-holding apparatus (Narishige) under a surgical stereoscope (Olympus). The mouse's body temperature was maintained at 37°C using a heating mat and rectal probe. The animal's head was shaved, and an eye-protecting ointment (Viscotears) was applied to both eyes. An incision in the scalp was made using surgical scissors, and the area was expanded with blunt dissection to expose the skull. The site for the craniotomy over S1 was identified based on the area of maximum tdTomato expression, detected using a custom-built LED (525 nm, 30 W). A tissue adhesive (Vetbond) was applied to fix the surrounding scalp to the skull and to secure cranial sutures. Multiple layers of dental cement (Simplex Rapid) were applied to create a chamber on top of the skull. A 0.5 mm craniotomy was drilled over the marked region using a dental drill (Foredom). The craniotomy was submerged in the cortex buffer [containing the following (in mM): 125 NaCl, 5 KCl, 10 HEPES, 2 MgSO₄·7H₂O, 2 CaCl₂·2H₂O, and 10 glucose]. The bone flap and dura were removed, and the animal was then relocated onto the recording apparatus. Neurons were visualized using a custom-built two-photon microscope, which consisted of a modified confocal scan unit (Olympus FV300) coupled to a Ti:sapphire laser (Newport Spectra-Physics Mai Tai HP; 930 nm, 30 mW), and images were acquired using the FluoView software (Olympus).

In vivo patch pipettes were prepared from borosilicate glass capillaries (Harvard Apparatus) using a vertical puller (Narishige) to obtain a long shaft and a tip size of 5–8 MΩ (Jouhanneau and Poulet, 2019). Patch pipettes were filled with the same internal solutions described above, with the addition of a 15 mM Alexa Fluor 594 dye (Thermo Fisher) and mounted into an Optopatcher pipette holder (A-M Sytems) on a micromanipulator (Sutter). stGtACR2 was activated via a 50 μm diameter optic fiber (Thorlabs) embedded within the Optopatcher holder and connected to a 473 nm laser (MBL-FN-473–150mW, CNI Laser). A silver Cl⁻ ground electrode (Multi Channel Systems) was lowered onto the surface of the nearby skull. The recording site was submerged in the cortex buffer. Pipettes with positive pressure (200–300 mBar) were lowered onto the brain surface under 4× magnification (Plan N, NA 0.10, Olympus). The pipette was then visualized under 40× magnification (LUMPlanFI/IR NA 0.80, Olympus) and inserted into the brain to a depth of 200–300 μm, and the positive pressure was first lowered to 150–170 mBar and then to 20–30 mBar when the pipette was close to the target neuron. The Alexa dye enabled the pipette to be located under the two-photon microscope (820 nm, 50 mW), and the positive pressure released the Alexa dye into the extracellular space, revealing “shadows” of neuronal cell bodies (Kitamura et al., 2008). A neuron was approached, and when consistent changes in electrode resistance appeared and the dimple of the fluorescent dye on the neuron’s membrane was visible, the positive pressure was released, and the holding potential was clamped at −70 mV in voltage-clamp mode to assist gigaseal formation. Other conventions were the same as for in vitro electrophysiological recordings.

Determining inhibitory reversal potentials

Voltage-clamp recordings were used to measure the reversal potential of stGtACR2 (E_stGtACR2) and the GABA_AR (E_GABAAR). Step protocols involved delivering a series of voltage steps to the neuron, during which a membrane conductance was evoked by either light activation of stGtACR2 or agonist activation of GABA_ARs. From a holding potential of −70 mV, the voltage was stepped in 10 mV increments from −130 to −30 mV, each step lasted 500 ms, with 10 s between steps, and conductances were evoked 100 ms following the start of the steps. For analysis purposes, the membrane current was measured immediately before (i.e., “baseline” current), and then during, the evoked conductance. The two currents were then superimposed on current–voltage (IV) plots. The voltage was corrected for series resistance effects (Traynelis, 1998), by calculating the voltage drop associated with the series resistance and subtracting this from the command voltage. The point at which the two IV curves intersect reflected the reversal potential of the evoked current. In addition, the baseline current was used to infer the resting membrane potential (RMP), defined as the membrane potential at which the fitted line was equal to zero. Voltage ramp protocols were used to monitor dynamic changes in E_stGtACR2. These ramp protocols afforded frequent measurements of E_stGtACR2, which could be made before and immediately following a Cl⁻ load. From a holding potential of −70 mV, neurons were subjected to voltage ramps under baseline conditions (“baseline ramp”) or during light activation of stGtACR2. Each voltage ramp lasted 150 ms and extended from 60 mV below the holding potential to 40 mV above the holding potential (i.e., from −130 to −30 mV, at a rate of 0.7 mV/ms). For analysis purposes, the first and last 15 ms of each ramp were always excluded to avoid transient currents associated with the pipette capacitance. In some recordings, voltage-gated sodium conductances were activated as the ramp reached its most depolarized levels. Such voltage-activated conductances were as likely to occur during ramps with or without light activation and were simply excluded from the analysis by avoiding the final portion of the ramp. To isolate the stGtACR2 current associated with each voltage ramp, the baseline ramp current (i.e., without light) was subtracted from the stGtACR2-associated ramp current (i.e., with light). The isolated stGtACR2 current was then plotted on an IV curve, where the voltage was corrected for series resistance effects, as above. E_stGtACR2 was defined as the voltage at which the isolated stGtACR2 current was equal to zero, and the amplitude of the stGtACR2 conductance was determined from the gradient of the IV curve. Gramicidin-perforated patch-clamp recordings were used to measure resting E_Inh, and whole-cell patch clamp recordings were used to measure time constants of E_Inh recovery following a Cl⁻ load. While the pipette Cl⁻ concentration would be expected to influence absolute E_Inh values in the whole-cell configuration, our estimates of E_Inh recovery dynamics assumed that any Cl⁻ dialysis from the pipette and diffusion within the cytosol remained stable during a recording, such that time constants of E_Inh recovery reflected transmembrane Cl⁻ extrusion mechanisms (Deisz et al., 2011). Neurons were subjected to a Cl⁻ load by combining a prolonged activation of stGtACR2 (1 s light pulse) and increasing the driving force for Cl⁻ to enter the neuron by clamping the membrane potential at −20 mV.

Single-cell RNA sequencing meta-analysis

The levels of KCC2 (SLC12A5) RNA expression in pyramidal neurons and PV interneurons was acquired from published single-cell RNA sequencing datasets from the mouse sensory cortex (GSE60361, GSE71585, GSE60361). Each dataset was first reorganized to capture the following variables: single-cell ID, cell type for each single-cell ID, gene tested, and a matrix of the RNA expression-level measurements corresponding to each single-cell and gene tested. To normalize the data, the raw KCC2 RNA count for each neuron was divided by the total expression level of all measured genes in that neuron, and the resulting number was multiplied by the mean gene expression level across all cells within the same dataset. The resulting data was then analyzed using the ARTool to perform post hoc-aligned rank transform to measure pairwise comparisons (Wobbrock et al., 2011; Elkin et al., 2021).

Experimental design and statistical analyses

Most of the statistical analyses was performed using the Python SciPy library (version 3.10.9, RRID:SCR_008394). The only exception was the analysis of time constants which was performed using the one-phase decay analysis tool in the GraphPad Prism software package (v8.0.0, Dotmatics, RRID:SCR_002798). In all analyses, an alpha value (p-value, p) of <0.05 was considered significant. All data is reported as mean ± standard error of the mean (SEM). For comparison statistics, appropriate parametric and nonparametric tests were used. The statistical design for each experiment is found in the Results section and corresponding figure legends.

Data and code availability

Patch-clamp data was analyzed using Python 3.10 using the pyABF package (Harden, 2022). All material required to reanalyze the data is available from the corresponding authors upon reasonable request.

Results

Using light-gated stGtACR2 to determine inhibitory reversal potentials

A light-gated anion channel, stGtACR2, can be used to elicit inhibitory photocurrents at the cell soma (Mahn et al., 2018), and its permeability follows the lyotropic series characteristic of other anion channels such as the GABA_AR (Govorunova et al., 2015; Sineshchekov et al., 2015). This suggests that these channels share a similar relative permeability to Cl⁻ and HCO₃⁻, which is supported by evidence that the reversal potential for GtACR2 approximates that of the GABA_AR (Messier et al., 2018). This generates the opportunity to develop an optogenetic strategy for determining a neuron's inhibitory reversal potential (E_Inh; Fig. 1A). To establish the feasibility of determining E_Inh optically, we expressed stGtACR2 in L2/3 pyramidal neurons of the mouse cortex. AAV encoding Cre-dependent stGtACR2 fused to a red fluorescent protein, FusionRed, was injected into S1 of CaMKIIα-Cre mice, at 4 weeks of age (Fig. 1B). Having confirmed that this resulted in the expression of stGtACR2-FusionRed in Cux2-expressing L2/3 pyramidal neurons (Fig. 1C), acute brain slices were prepared 2 weeks after AAV injections, and stGtACR2-expressing L2/3 pyramidal neurons were targeted for whole-cell voltage-clamp and patch-clamp recordings (Fig. 1D). To evoke stGtACR2 photocurrents, brief pulses of blue light (470 nm laser) were targeted at the recorded neuron. Meanwhile, to generate an independent estimate of E_Inh in the same neuron, we used a traditional approach of delivering focal puffs of GABA to activate GABA_ARs. The agonist-evoked GABA_AR currents were pharmacologically isolated by blocking GABA_BRs with CGP55845 (10 μM).

Figure 1.

Using light-gated stGtACR2 to determine inhibitory reversal potentials. A, Schematic showing transmembrane Cl⁻ fluxes elicited by GABA_AR agonist activation. An alternative, optogenetic strategy elicits transmembrane Cl⁻ fluxes via light activation of stGtACR2. B, Cre-dependent stGtACR2-FusionRed was delivered by intracerebral AAV injection into the S1 of CaMKIIα-Cre mice. C, Confocal image showing stGtACR2-FusionRed expression in Cux2⁺ L2/3 pyramidal neurons at 6 weeks of age, 2 weeks after injection. D, Experimental setup in which whole-cell patch-clamp recordings were performed from stGtACR2-expressing L2/3 pyramidal neurons. stGtACR2 was activated with blue light pulses (470 nm laser) and delivered via the microscope objective. Focal applications of exogenous GABA from a separate pipette were used to activate GABA_ARs. Inset, current-clamp recording from an example stGtACR2-expresssing regular spiking pyramidal neuron. E, IV plot from a voltage-clamp-step protocol shows the stGtACR2-induced response (cyan, 100 ms light) and baseline current (black) at different holding potentials. The reversal potential for both stGtACR2 (E_stGtACR2) and the baseline current (equivalent to the RMP) are indicated. Inset shows raw traces and indicates where baseline and stGtACR2 currents were measured (dashed vertical blue lines). F, IV plot from the same neuron as in “E” showing the GABA-evoked response (purple) and baseline current (black) at different holding potentials, from which E_GABAAR and RMP were determined.

The reversal potential for stGtACR2 (E_stGtACR2) was calculated by measuring the amplitude of the light-gated stGtACR2 photocurrent elicited while clamping the neuron at different holding potentials (Fig. 1E). This enabled us to generate separate IV plots for the baseline holding current and for the total holding current recorded during stGtACR2 activation, whose point of intersection represents E_stGtACR2 (Fig. 1E). Meanwhile, an independent estimate of E_Inh was made by measuring the amplitude of agonist-evoked GABA_AR currents elicited while clamping at different holding potentials and inferring the reversal potential of current flowing through the GABA_AR (E_GABAAR) in an analogous manner (Fig. 1F). These recordings revealed similar estimates of E_Inh in the same neuron (Fig. 1E,F). To explore whether the relationship between E_stGtACR2 and E_GABAAR was maintained across a wide range of intraneuronal Cl⁻ concentrations, a series of recordings were performed using one of the three different intracellular pipette solutions (4, 20, or 70 mM Cl⁻), such that the intraneuronal Cl⁻ concentration was altered by dialysis. Across a population of neurons (n = 24 neurons from seven mice), E_stGtACR2 and E_GABAAR were found to be significantly correlated with one another (p < 0.0001, Pearson’s correlation coefficient; Fig. 2), supporting the idea that E_stGtACR2 can be used to estimate somatic E_Inh.

Figure 2.

stGtACR2 reports transmembrane inhibitory reversal potentials. Across a population of neurons using a wide range of intracellular pipette Cl⁻ concentrations, the reversal potential of stGtACR2 (E_stGtACR2) and the reversal potential of agonist-activated GABA_ARs (E_GABAAR) were highly correlated (linear fitted line, black solid line, r = 0.99, n = 24 neurons from 7 mice, p < 0.0001, Pearson’s correlation coefficient) and approached equality (dashed black line).

In vivo optogenetic measurements of E_Inh

A significant advantage of determining E_Inh optically is that it eliminates the requirement to deliver a receptor agonist, which can be technically challenging, particularly in intact tissue. To illustrate this, we asked whether stGtACR2 could be used to determine E_Inh in vivo in the intact mouse cortex, by combining two-photon-targeted recordings with optical activation of stGtACR2. To optimize the tissue for two-photon imaging, IUE was used to express stGtACR2 in L2/3 pyramidal neurons of the S1 (Fig. 3A). The cytosolic fluorescent reporter protein tdTomato was coexpressed to aid in the screening of the resulting pups (see Materials and Methods) and to facilitate identification of the transfected neurons in the mature cortex (Fig. 3B). Once the animals had reached 6 weeks of age, they were anesthetized with urethane, and two-photon-targeted in vivo voltage-clamp and patch-clamp recordings were performed from stGtACR2-expressing L2/3 pyramidal neurons (Fig. 3C). Alexa 594 dye was added to the patch pipette solution to visualize the pipette under 820 nm light, while tdTomato was visualized under 930 nm light (Fig. 3D).

Figure 3.

Optogenetic determination of E_Inh in vivo. A, IUE was used to coexpress stGtACR2 and tdTomato in L2/3 pyramidal neurons of the mouse S1. B, Confocal image from the S1 at 6 weeks of age. tdTomato facilitated identification of expressing neurons. C, Experimental setup for two-photon-targeted patch-clamp recordings. L2/3 pyramidal neurons coexpressing stGtACR2 and tdTomato were visualized at 930 nm. Patch pipettes containing Alexa 594 dye were visualized at 820 nm. Light pulses were delivered via an optic fiber inserted within the patch pipette. D, In vivo two-photon image of tdTomato-expressing L2/3 pyramidal neurons. E, Example image of an in vivo whole-cell patch-clamp recording (pipette solution, 4 mM Cl⁻). The Alexa 594 dye (magenta) diffuses from the pipette (white lines) into the recorded L2/3 pyramidal neuron. F, IV plot from in vivo voltage-step protocol in whole-cell mode shows stGtACR2-induced response (cyan) and baseline current (black) at different holding potentials. Inset shows raw traces and indicates where baseline and stGtACR2 currents were measured (dashed vertical blue lines). The reversal potential for stGtACR2 (E_stGtACR2) and the baseline current (RMP) are indicated. G, Example image of an in vivo gramicidin (Gx)-perforated patch-clamp recording from a L2/3 pyramidal neuron. Alexa 594 dye does not enter the neuron due to the perforated status of the patch. H, IV plot from in vivo gramicidin voltage-step protocol shows stGtACR2-induced response (cyan) and baseline current (black) at different holding potentials.

Initially, recordings performed in whole-cell configuration (Fig. 3E) confirmed that it was possible to combine light activation of stGtACR2 currents with voltage-step protocols at different holding potentials, thereby generating estimates of E_stGtACR2 in vivo (Fig. 3F). Subsequent recordings incorporated a cation-selective perforating agent, gramicidin, into the patch pipette solution, so as to preserve the neuron's native transmembrane Cl⁻ gradients (Fig. 3G). We combined our published protocol for performing gramicidin-perforated patch-clamp recordings in vivo (Alfonsa et al., 2023; Burman et al., 2023) with two-photon targeting of stGtACR2-expressing neurons. Perforation with gramicidin was confirmed by monitoring the lack of Alexa 594 dye diffusion into the neuron and from a decrease and then stabilization in series resistance following gigaseal formation (Fig. 3G). These recordings revealed that it was also possible to estimate resting E_Inh under undisturbed conditions in vivo, by combining light-gated stGtACR2 currents with voltage-step protocols (Fig. 3H).

Optogenetic determination of E_Inh dynamics

Transmembrane Cl⁻ fluxes can alter [Cl⁻]_i, meaning that rather than being static, E_Inh is a dynamic parameter that reflects the recent history of the cell. For example, Cl⁻ influxes into neurons via GABA_ARs can cause transient increases in [Cl⁻]_i, which lead to depolarizing shifts in E_Inh (Staley and Proctor, 1999; Jin et al., 2005; Raimondo et al., 2012). The rate at which E_Inh subsequently recovers is reflective of the neuron's capacity to extrude Cl⁻ and is an important measure of a neuron's capacity to regulate its [Cl⁻]_i (Deisz et al., 2011; Raimondo et al., 2012; Doyon et al., 2016a; Lanz et al., 2020; Otsu et al., 2020). An optogenetic strategy for quantifying E_Inh dynamics has the potential to avoid confounds associated with agonist-based techniques, including controlling the amount of agonist, activating different receptor pools, and agonist-induced changes to receptors (Fig. 4A). Furthermore, by capitalizing on the temporal control afforded by light-gated channels, there is the potential to quantify a cell's E_Inh dynamics and Cl⁻ extrusion processes in real time, rather than having to infer across separate agonist applications. To investigate this possibility, we performed whole-cell patch-clamp recordings from stGtACR2-expressing L2/3 pyramidal neurons in acute brain slices (4 mM Cl⁻ pipette concentration; Fig. 4B). While the whole-cell configuration would be expected to influence the neuron's absolute E_Inh, our intention was to combine the recordings with pharmacological methods to estimate the contribution that KCC2 makes to the neuron's Cl⁻ extrusion following a Cl⁻ load. We first asked whether stGtACR2 activation can induce an intracellular Cl⁻ load that causes a temporary depolarizing shift in E_Inh. Adopting a protocol similar to that used by Raimondo et al. (2012), we activated stGtACR2 continuously for 1 s while depolarizing the membrane potential to increase the driving force for Cl⁻ to enter the neuron (Fig. 4C). This resulted in a robust and transient Cl⁻ load, as brief stGtACR2 photocurrents switched in polarity, such that the outward currents that were observed before the optical Cl⁻ load became inward immediately after the Cl⁻ load, but then decreased in amplitude, and reversed to become outward again over a period of tens of seconds (Fig. 4C,D).

Figure 4.

Optogenetic determination of E_Inh dynamics. A, Schematic showing intracellular Cl⁻ being raised (“loading”) during stGtACR2 activation and then extruded by KCC2. B, Whole-cell patch-clamp recordings (pipette solution, 4 mM Cl⁻) were performed from stGtACR2-expressing L2/3 pyramidal neurons in acute brain slices from the S1. C, A neuron experienced a Cl⁻ load by pairing long stGtACR2 activation (1 s light pulse, “Load”) with membrane potential depolarization (to −20 mV), to increase the driving force for Cl⁻ to enter the neuron. Following the Cl⁻ load, delivering short light pulses (100 ms) every second revealed that the stGtACR2-current had flipped from being an outward current to transiently becoming an inward current, which then gradually returned to being an outward current. D, Data from the recording in “C” showing the overlaid stGtACR2-current responses after the Cl⁻ load (top) and a plot capturing the timescale of recovery of the stGtACR2-current responses (bottom). E, Voltage-clamp protocol used to monitor E_stGtACR2 dynamics relative to a stGtACR2-mediated Cl⁻ load. The traces represent the recorded current (top) and the estimated holding potential following series resistance correction (bottom). The L2/3 pyramidal neuron experienced a Cl⁻ load by pairing long stGtACR2 activation (1 s light pulse, “Load”) with membrane potential depolarization (to −20 mV). A series of voltage ramps (each 150 ms duration, labeled t₁–t₁₁) were then paired with a brief stGtACR2 activation (150 ms light pulse) to measure E_stGtACR2 dynamics at 5 s intervals over a period of 60 s. To isolate the stGtACR2 current, a baseline ramp current (without light) was subtracted from each stGtACR2-associated ramp current. F, IV plots from data in “E” showing the isolated stGtACR2 current at t₁, t₂, and t₁₁. The x-axis intercept represents E_stGtACR2. G, IV plot of isolated stGtACR2 currents from t₁ to t₁₁. Inset, plot of E_stGtACR2 recovery following the Cl⁻ load. A single exponential fit provides the time constant of Cl⁻ extrusion (τ_Cl-extrusion). H, The stGtACR2 conductance remained stable within each ramp (top; mean percentage change in conductance across 20 ms periods was −15.71 ± 4.01%, p = 0.83, one-way ANOVA) and across ramps (bottom; −10.71 ± 2.91% conductance change at intervals of 5 s, p = 0.86, one-way ANOVA), consistent with a modest contribution of desensitization mechanisms (n = 5 neurons from 3 mice). I, E_stGtACR2 recovery dynamics in L2/3 pyramidal neurons under control conditions (−VU; dark cyan) and when KCC2 was blocked with the selective antagonist VU0463271 (+VU, 10 μM; light cyan). Dashed line indicates the mean, and the shaded area indicates SEM. J, KCC2 block results in an increase in resting E_stGtACR2 (−VU: −78.86 ± 2.56 mV vs +VU: −70.12 ± 0.98 mV, p = 0.01, paired t test). K, KCC2 block results in a slowing in Cl⁻ extrusion (−VU: 7.29 ± 0.41 s vs +VU: 15.79 ± 0.98 s, p = 0.001, paired t test). *, p < 0.05. **, p < 0.01.

Having established that stGtACR2 can induce Cl⁻ loads, we designed an optogenetic protocol to quantify E_Inh dynamics and assess the contribution of KCC2 to Cl⁻ extrusion (Fig. 4E). To generate frequent measurements of E_stGtACR2 before and immediately following a Cl⁻ load, IV plots of stGtACR2 photocurrents were produced by subtracting the current response to a control voltage ramp (“baseline ramp”) from the response to the same voltage ramp delivered during stGtACR2 activation with blue light (Fig. 4E,F). These repeated measurements revealed that E_stGtACR2 had become more depolarized immediately following an optical Cl⁻ load and recovered to hyperpolarized values over a period of seconds. From these real-time dynamic measurements, the neuron's Cl⁻ extrusion capacity could be defined as the time constant of E_stGtACR2 recovery dynamics (τ_Cl-extrusion; Fig. 4G), which was 7.29 ± 0.41 s under control conditions (n = 5 neurons from three mice). We noted that the stGtACR2 conductance was stable both within each ramp and between ramps, consistent with low levels of stGtACR2 desensitization during these measurements (Fig. 4H). To determine the contribution of KCC2 to these E_Cl dynamics, we added the KCC2-specifc antagonist VU0463271 (“VU”) to the perfusate and repeated the protocol (Fig. 4I). In addition to an effect upon resting E_stGtACR2 values (Fig. 4J), blocking KCC2 resulted in a slowing in the rate of E_stGtACR2 recovery dynamics, such that the τ_Cl-extrusion more than doubled (Fig. 4K). Thus, our measurements of E_stGtACR2 provide an effective readout of E_Inh dynamics while avoiding the confounders associated with agonist-based techniques.

Optogenetic discrimination of neuron-specific differences in E_Inh

Optimal methods for determining E_Inh should be able to distinguish cells that differ in their ion handling mechanisms. Several reports have suggested that parvalbumin (PV)-expressing fast-spiking interneurons and pyramidal neurons exhibit differences in their Cl⁻ homeostasis, although there has been no consensus on the relative contribution that cation Cl⁻ cotransporters (CCCs) make in these neuron types (Verheugen et al., 1999; Martina et al., 2001; Banke and McBain, 2006; Vida et al., 2006; Sauer and Bartos, 2010; Otsu et al., 2020). We took advantage of the Cre-lox recombination system and our optogenetic approach to investigate potential differences in resting E_Inh and E_Inh dynamics between PV interneurons and pyramidal neurons. We performed intracortical viral injections of Cre-dependent stGtACR2-FusionRed into S1 of PV-Cre transgenic mice (Fig. 5A). Using this approach, we were able to restrict stGtACR2 expression to PV interneurons (Fig. 5B), which could then be targeted for gramicidin-perforated patch-clamp recordings, to preserve native transmembrane Cl⁻ gradients (Fig. 5C). Equivalent recordings were performed from pyramidal neurons following stGtACR2 expression in the S1 of CaMKIIα-Cre mice (as in Fig. 1B).

Figure 5.

Optogenetic discrimination of neuron-specific differences in resting E_Inh. A, Mice expressing Cre recombinase in PV interneurons received S1 injections of Cre-dependent stGtACR2-FusionRed AAV. B, Confocal image showing selective expression of stGtACR2-FusionRed (red) in L2/3 immunohistochemically confirmed PV interneurons (green). C, Setup in which gramicidin (Gx)-perforated patch-clamp recordings were performed from stGtACR2-expressing PV interneurons. D, Resting E_stGtACR2 was measured in PV interneurons (turquoise, top) and pyramidal neurons (blue, bottom) using a voltage-clamp-step protocol. The resulting IV plots indicate resting E_stGtACR2 under control conditions (−VU; darker) and when KCC2 was blocked with VU0463271 (+VU; lighter). E, Resting E_stGtACR2 was more depolarized in PV interneurons compared with pyramidal neurons (PV: −69.16 ± 4.54 mV, n = 6 neurons from 3 mice vs Pyr: −77.12 ± 3.83 mV, n = 7 neurons from 5 mice, p = 0.006, unpaired t test). Pyramidal neuron data replotted from Figure 4. F, The RMP of PV interneurons was more depolarized than the RMP of pyramidal neurons (PV: −60.74 ± 1.26 mV vs Pyr: −72.68 ± 1.31 mV, p < 0.0001, unpaired t test). G, The driving force for stGtACR2 (DF_stGtACR2= RMP − E_stGtACR2) was not significantly different between the cell types (PV: 8.42 ± 1.71 mV vs Pyr: 4.50 ± 1.82 mV, p = 0.15, unpaired t test). H, PV interneurons and pyramidal neurons both exhibited a significant contribution of KCC2 to resting E_stGtACR2 (PV: 9.79 ± 1.62 mV, n = 6 neurons from 3 mice, p = 0.002, one-sample t test; Pyr: 12.92 ± 1.76 mV, n = 7 neurons from 5 mice, p = 0.0003, one-sample t test), and there was no detectable difference in the contribution of KCC2 to resting E_stGtACR2 in the two neuronal populations (p = 0.23, unpaired t test). **, p < 0.01; ***, p < 0.001; ns, not significant.

Voltage-step protocols were used to compare resting E_stGtACR2 in L2/3 PV interneurons and pyramidal neurons, before and after pharmacologically blocking KCC2 with VU (Fig. 5D). This revealed a more depolarized resting E_stGtACR2 in PV interneurons than pyramidal neurons (Fig. 5E), confirming that optogenetic E_Inh measurements can distinguish between neuron types. The PV interneurons also had more depolarized RMPs than the pyramidal neurons (Fig. 5F), and, consequently, there was no difference in the driving force for stGtACR2 (E_stGtACR2 − RMP) between the two cell types at rest (Fig. 5G). VU application caused a depolarizing shift in the resting E_stGtACR2 of PV interneurons and pyramidal neurons, revealing that KCC2-mediated extrusion contributes to resting E_Inh in both neuron types (Fig. 5D). The VU-associated shifts in resting E_stGtACR2 were comparable for the two neuron types, suggesting that the neuron types could not be distinguished in terms of KCC2's contribution (Fig. 5H).

Optogenetic E_Inh dynamics reveal neuron-specific KCC2-mediated extrusion

It has been suggested that KCC2 activity is better revealed when a neuron is subjected to a Cl⁻ load, as this challenges the neuron's capacity to extrude Cl⁻ (Jarolimek et al., 1999; Payne et al., 2003; Deisz et al., 2011). To test if optogenetic E_Inh measurements can be used to resolve more subtle cell-type differences in KCC2 activity, we used whole-cell patch-clamp recordings to quantify E_stGtACR2 dynamics in L2/3 PV interneurons and compared these with equivalent recordings in pyramidal neurons (Fig. 6A). As we had observed in pyramidal neurons (Fig. 4), E_stGtACR2 in PV interneurons exhibited depolarized values immediately following a stGtACR2-mediated Cl⁻ load and then gradually returned to hyperpolarized values over a period of seconds (Fig. 6B,C). The amplitude of the PV interneuron stGtACR2 conductance was stable within each ramp, and between ramps, consistent with a modest contribution of desensitization mechanisms during the estimates of PV interneuron E_stGtACR2 dynamics (Fig. 6D). Each PV interneuron's Cl⁻ extrusion capacity was therefore defined from the time constant of E_stGtACR2 recovery dynamics (τ_Cl-extrusion; Fig. 6C), which was shown at the population level to slow when KCC2 was blocked with VU (Fig. 6E).

Figure 6.

Optogenetic E_Inh dynamics reveal neuron-specific differences in KCC2-mediated extrusion. A, Whole-cell patch-clamp recordings (pipette solution, 4 mM Cl⁻) were performed from stGtACR2-expressing L2/3 PV interneurons in the S1. Inset, current-clamp recording from an example stGtACR2-expresssing fast-spiking PV interneuron. B, Voltage-clamp ramp protocol used to monitor E_stGtACR2 recovery following a stGtACR2-mediated Cl⁻ load. Example data from a PV interneuron indicate the recorded current (top), estimated holding potential following series resistance correction (middle), and resulting IV plots (bottom) showing the isolated stGtACR2 current at t₁, t₂, and t₁₁. Conventions as in Figure 4. C, IV plot of isolated stGtACR2 currents from t₁ to t₁₁ in a PV interneuron. Inset, E_stGtACR2 recovery following Cl⁻ load. A single exponential fit provides the time constant of Cl⁻ extrusion (τ_Cl-extrusion). D, The amplitude of the PV interneuron stGtACR2 conductance remained stable within each ramp (top; mean percentage change in conductance across 20 ms periods was −6.26 ± 2.98%, p = 0.24, one-way ANOVA) and across ramps (bottom; −8.51 ± 2.42% conductance change at intervals of 5 s, p = 0.82, one-way ANOVA), consistent with a modest contribution of desensitization mechanisms (n = 6 neurons from 3 mice). E, E_stGtACR2 recovery in PV interneurons under control conditions (−VU; dark cyan) and when KCC2 was blocked with the selective antagonist VU0463271 (+VU, 10 μM; light cyan). Inset, KCC2 block results in a slowing in Cl⁻ extrusion in PV interneurons (−VU: 11.38 ± 0.65 s vs +VU: 15.57 ± 0.63 s, p = 0.006, paired t test). F, E_stGtACR2 recovery following a Cl⁻ load was slower in PV interneurons than pyramidal neurons. Inset, Cl⁻ extrusion time constant (τ_Cl-extrusion) was greater in PV interneurons (PV: 11.38 ± 0.65 s vs Pyr: 7.29 ± 0.41 s, p = 0.0007, unpaired t test). Pyramidal neuron data replotted from Figure 4. G, E_stGtACR2 recovery and the resulting Cl⁻ extrusion time constant were indistinguishable in PV interneurons and pyramidal neurons after KCC2 block (PV: 15.57 ± 0.63 s vs Pyr: 15.79 ± 0.98 s, p = 0.84, unpaired t test). H, While PV interneurons and pyramidal neurons both exhibited a significant contribution of KCC2 to their Cl⁻ extrusion time constant (PV: p = 0.004, one-sample t test; Pyr: p = 0.001, one-sample t test), KCC2's contribution to Cl⁻ extrusion was significantly less in PV interneurons (PV: 4.13 ± 0.79, n = 6 neurons from 3 mice vs Pyr: 8.01 ± 0.92, n = 5 neurons from 3 mice, p = 0.01, unpaired t test). I, Meta-analysis of single-cell RNA sequencing data on KCC2 levels in isolated pyramidal neurons and PV interneurons from the sensory cortex. Normalized KCC2 RNA expression levels are lower in PV interneurons than pyramidal neurons in three separate studies: (i) Zeisel et al. (2015) in the S1 (PV: 2.54 ± 0.62, n = 10 neurons vs Pyr: 5.23 ± 0.23, n = 286 neurons, p = 0.07, Kolmogorov–Smirnov test, Dunn's multiple comparison test), (ii) Tasic et al. (2016) in the V1 (PV: 9.73 ± 0.59, n = 270 neurons vs Pyr: 16.40 ± 0.71, n = 756 neurons, p = 0.0073, Kolmogorov–Smirnov test), and (iii) Tasic et al. (2018) in the V1 (PV: 209.5 ± 3.25, n = 1,337 neurons vs Pyr: 264.3 ± 2.94, n = 7,366 neurons, p < 0.0001, Kolmogorov–Smirnov test). *, p < 0.05; **, p < 0.01; ***, p < 0.001.

A direct comparison of these dynamics across cell types revealed clear differences between PV interneurons and pyramidal neurons (Fig. 6F–H). Firstly, the rate of E_stGtACR2 recovery was slower in PV interneurons than in pyramidal neurons (Fig. 6F). This cell-type difference was abolished when KCC2 was blocked with VU (Fig. 6G), and a direct comparison of KCC2's contribution with τ_Cl-extrusion revealed that KCC2 makes a smaller contribution to Cl⁻ extrusion mechanisms in PV interneurons (Fig. 6H). To corroborate these functional observations, we performed a meta-analysis of single-cell RNA sequencing studies that have distinguished pyramidal neurons and PV interneurons in the mouse sensory cortex (Zeisel et al., 2015; Tasic et al., 2016, 2018; Fig. 6I). After normalizing for total gene expression per cell, we observed that the expression levels of KCC2 were lower in PV interneurons compared with pyramidal neurons. This difference was evident in two independent large-scale datasets from the primary visual cortex (V1), and a similar trend was also observed in a dataset from the S1 that comprised a smaller number of PV interneurons. Taken together, these results support the conclusion that optogenetic estimates of E_Inh dynamics can reveal functionally relevant differences in neuron-specific Cl⁻ extrusion mechanisms.

Discussion

Here, we describe an optogenetic strategy for probing resting and dynamic E_Inh in a cell-specific and subcellular-targeted manner. We demonstrate that this meets important prerequisites for studying inhibitory reversal potentials while offering advantages over traditional agonist-based methods. The strategy is validated in vitro and in vivo, is shown to capture differences in E_Inh dynamics following manipulations of endogenous Cl⁻ fluxes, and reveals distinct resting E_Inh across genetically defined neuronal subpopulations. Furthermore, using our optogenetic approach to challenge Cl⁻ homeostasis, we quantify a cell's E_Inh dynamics and uncover cell-specific Cl⁻ extrusion processes that are supported by the differential expression of endogenous handling mechanisms.

Our characterization focused on the soma-targeted, anion channel, stGtACR2 (Govorunova et al., 2015; Alfonsa et al., 2016; Mahn et al., 2018). To establish the potential of using stGtACR2 to estimate E_Inh, E_stGtACR2 was determined across a range of intracellular Cl⁻ concentrations (4–70 mM) and compared with estimates of E_GABAAR. The strong, positive correlation across a population of neurons supported the conclusion that E_stGtACR2 reflects E_Inh. This is consistent with evidence that GtACR2 permeability follows the lyotropic series characteristic of other anion channels and suggests a similar relative permeability to Cl⁻ and HCO₃⁻ as the GABA_AR, although this remains to be tested directly (Bormann et al., 1987; Govorunova et al., 2015; Sineshchekov et al., 2015; Messier et al., 2018). Further validation came from our gramicidin-perforated recordings of resting E_stGtACR2 and whole-cell recordings of E_stGtACR2 recovery following a Cl⁻ load. Our estimated time constant for E_stGtACR2 recovery in pyramidal neurons (∼7.3 s) was close to previous estimates of the time constant for E_Cl recovery (∼8 s across six separate studies; Thompson et al., 1988; Luhmann and Prince, 1991; Staley and Proctor, 1999; Jin et al., 2005; Deisz et al., 2011; Raimondo et al., 2012) and is a parameter that has been shown to be invariant to the size of Cl⁻ load (Deisz et al., 2011). Finally, we also confirmed in multiple cell types that both resting E_stGtACR2 and E_stGtACR2 recovery are sensitive to manipulations of a Cl⁻ exporter, KCC2 (Misgeld et al., 1986; Thompson et al., 1988; Jarolimek et al., 1999; DeFazio et al., 2000).

Although we use the term E_Inh, the reality is that our optogenetic strategy can estimate the reversal potential regardless of whether the net current flowing through stGtACR2 is outward or inward. The strategy will therefore be effective under conditions when ligand-gated anion channels (e.g., GABA_A or glycine receptors) exert either a hyperpolarizing or depolarizing effect relative to the RMP. This is relevant during brain development, when E_Cl is often more depolarized, and an optogenetic strategy may be useful in tracking the maturation of inhibitory synaptic systems (Akerman and Cline, 2007; Richards et al., 2010; Kaila et al., 2014; van Rheede et al., 2015). Similarly, relatively depolarized E_Cl values have been observed in a variety of pathological states associated with hyperexcitability, meaning that our optogenetic strategy could be useful in investigations of epilepsy and neuropathic pain (Coull et al., 2003; Kaila et al., 2014; Burman et al., 2022).

Estimates of E_Inh have traditionally been made by activating endogenous receptors, either by exogenous application of an agonist (e.g., focal delivery or uncaging of GABA) or by stimulating the release of a relevant endogenous agonist (e.g., electrical or optical activation of presynaptic GABA-releasing axons). There are, however, limitations associated with inferring resting and dynamic E_Inh from endogenous receptors (e.g., GABA_ARs). This includes the fact that repeated agonist delivery can alter the receptor's properties and availability via desensitization mechanisms (Frosch et al., 1992; Jones and Westbrook, 1995; Gravielle, 2018). Meanwhile, when stimulating GABA-releasing axons, presynaptic mechanisms can affect the probability of neurotransmitter release, which could complicate the interpretation of dynamic E_Inh measurements and manipulations that affect the presynaptic neurons (Kullmann et al., 2005). Furthermore, although GABA_ARs are thought to contribute to a neuron's E_Inh (Karlsson et al., 2011; Yelhekar et al., 2016; Burman et al., 2023), it is difficult to measure their contribution (e.g., by pharmacological blockade) if the GABA_AR is also being used to estimate E_Inh.

The stGtACR2-based optogenetic approach addresses these potential limitations associated with agonist-based methods, including avoiding the requirement to pharmacologically isolate an endogenous receptor response (Prenosil et al., 2006; Pamenter et al., 2011). Indeed, the optogenetic approach could be used in cases where a cell type lacks ligand-gated anion channels, where delivering an agonist would not be an option. The ability to restrict Cl⁻ loads to a defined period of light activation distinguishes our approach from techniques that involve the application of a continuous intracellular Cl⁻ load via a whole-cell patch pipette, for example (Jarolimek et al., 1999). A further advantage is that optogenetic approaches lend themselves to studies in intact preparations. Accordingly, we show that stGtACR2 can be used to determine E_Inh in defined neurons of the intact brain, generating the potential for experiments that could combine dynamic E_Inh measurements and manipulations of Cl⁻-regulatory processes in vivo. This application represents an important complement to the use of Cl⁻ imaging approaches, which can provide information on cytosolic Cl⁻ concentration in vivo, but do not capture E_Inh (Sulis Sato et al., 2017; Boffi et al., 2018; Rahmati et al., 2021).

The GtACR2-based optogenetic approach offers excellent control over the spatial and temporal resolution and amplitude of the evoked anion currents. The spatial resolution is determined in part by GtACR2's localization within the cell of interest and in part by the resolution of the optical system. This is attractive, as E_Inh is thought to vary between subcellular compartments (Duebel et al., 2006; Szabadics et al., 2006; Földy et al., 2010; Ellender et al., 2014). One could imagine examining E_Inh in dendrites or at the axon initial segment by targeting GtACR2 to the relevant subcellular compartment and/or using more focal light delivery. Light-controlled methods also make it easy to vary the duration and intensity of the anion current, which allows the experimenter to systematically vary Cl⁻ fluxes over a range of durations (milliseconds to tens of seconds) and amplitudes (via light intensity). This can help to minimize unwanted changes to E_Inh caused by sustained endogenous receptor activation (Ehrlich et al., 1999; Akerman and Cline, 2006) while allowing one to control the duration and intensity of Cl⁻ fluxes in order to quantify E_Inh dynamics.

By targeting stGtACR2 to genetically defined cell types, we demonstrated that PV interneurons and pyramidal neurons in L2/3 of the mouse visual cortex differ in their resting E_stGtACR2. This supports previous evidence of more depolarized E_Inh values in interneurons of the mature hippocampus and cortex (Martina et al., 2001; Otsu et al., 2020; Călin et al., 2023). When related to cell-type differences in RMPs, this can generate similar GABA_AR driving forces under resting conditions (Otsu et al., 2020), although there are likely to be differences related to age and the level of network activity (Sauer and Bartos, 2010; Burman et al., 2023). Indeed, such differences in somatic E_Inh may underlie cell-type-specific functions. For example, it has been proposed that relatively depolarized E_Inh values in PV interneurons may tend to favor shunting inhibition and enable PV interneurons to promote coherent oscillatory network activity (Vida et al., 2006). An interesting future direction will be to use optogenetic approaches to determine E_Inh dynamics in different, genetically defined, cell populations.

In terms of the mechanisms that determine a cell's transmembrane gradients, we found that KCC2 makes a similar contribution to resting E_Inh in both PV interneurons and pyramidal neurons, supporting recent evidence from the hippocampus (Otsu et al., 2020). This would suggest that KCC2 is not responsible for the different resting E_Inh between these two neuron types and that other ion fluxes contribute. The biophysical prediction is that more subtle differences in a cotransporter's activity can be revealed under an ionic load, which might better represent a neuron's state within an active network (Doyon et al., 2016b; Burman et al., 2023). In line with this prediction, we capitalized upon the control of stGtACR2 to both challenge and then monitor E_Inh recovery through rapid and sequential measurements. This enabled us to quantify E_Inh dynamics and Cl⁻ extrusion processes in real time, which is difficult with agonist-based approaches (Raimondo et al., 2012). We reveal that PV interneurons exhibit a slower recovery to a transient Cl⁻ load and that this difference in E_Inh dynamics is due to a differential contribution by KCC2, as blocking the cotransporter normalized E_Inh dynamics across neurons. This was corroborated by our meta-analysis of single-cell RNA sequencing data from the mouse sensory cortex, which revealed higher KCC2 expression in pyramidal neurons than PV interneurons. We therefore propose that our optogenetic strategy generates sensitive estimates of E_Inh dynamics, avoiding issues associated with repeated activation of endogenous Cl⁻-permeable receptors.

In conclusion, we establish an optogenetic and agonist-independent strategy for probing E_Inh in vitro and in vivo, in a manner that is compatible with the targeting of defined cell populations and subcellular compartments. This enabled us to quantify the contribution of endogenous ion fluxes to E_Inh, to distinguish cell-type differences in resting E_Inh, and to reveal differences in E_Inh dynamics between neuronal subpopulations. More generally, this underscores the fact that optogenetic techniques afford opportunities beyond their principal application of modulating neuronal activity. Our findings therefore help to expand the repertoire of optogenetics and encourage the application of other light-activated ion channels in physiological investigations of transmembrane ion gradients and their associated reversal potentials.