Overexpression or Activation of Potassium Channel TASK-3 Protects Retinal Ganglion Cells and Restores Visual Function in Optic Nerve Crush

Abstract

Purpose

This study explored in a mouse model whether activation or upregulation of the two-pore domain potassium channel tandem pore domain acid-sensitive potassium channel 3 (TASK-3) in retinal ganglion cells (RGCs) could protect RGCs and reverse the vision loss arising through optic nerve injury.

Methods

TASK-3 activity was assessed using patch-clamp electrophysiology. The optic nerve of each mouse was crushed, and the selective TASK-3 agonist CHET3 was applied to the surface of the eye once daily for 1 week or TASK-3 was overexpressed specifically in RGCs through infection with recombinant adeno-associated virus 1 week after optic nerve crushing. Numbers of RGCs and of intrinsic photosensitive RGCs were determined through fluorescence microscopy. Image-forming activity of RGCs in mice was assessed using flash visual evoked potentials, the visual cliff test, and the visual water maze task. The non–image-forming activity of intrinsic photosensitive RGCs was assessed using the pupillary light reflex test.

Results

CHET3 treatment increased the number of RGCs surviving after optic nerve injury, and it improved their electrophysiological response, visual acuity, contrast sensitivity, and the sensitivity of pupillary light reflex. These effects were associated with decreased RGC excitability. TASK-3 overexpression in sparse RGCs surviving long-term optic nerve injury restored their image- and non–image-forming activities.

Conclusions

These results suggest that pharmacological activation or upregulation of TASK-3 may be a promising therapeutic strategy to promote vision recovery after optic nerve injury or in eye disorders associated with RGC degeneration.

Retinal ganglion cells (RGCs), located in the innermost layer of the retina, project axons that form the optic nerve, which projects visual signals onto the visual cortex in the brain.¹ Traumatic injury or diseases such as glaucoma or optic glioma can damage the optic nerve² and overexcite RGCs, leading to their irreversible loss and concomitant loss of vision.³ Approaches that protect the survival and activity of RGCs after optic nerve injury could be powerful therapies to stabilize and even restore visual function.^4–6

Previously, we showed that RGCs express the two-pore domain potassium channel tandem pore domain acid-sensitive potassium channel 3 (TASK-3), which regulates RGC excitability by sensing extracellular acidification.⁷ We further showed that, as mice age, levels of TASK-3 expression fall, leading to a decline in both visual acuity and contrast sensitivity, even though the number of RGCs remains unchanged. Overexpressing TASK-3 in the RGCs can reverse this age-related loss of visual function. These findings led us to hypothesize that activation or overexpression of TASK-3 might help rescue vision after optic nerve injury.

Here, we explored this possibility using a mouse model of optic nerve injury, which we treated with the selective TASK-3 agonist CHET3, a biguanide used against neuropathic pain.⁸ Animals were also infected with adeno-associated virus (AAV) encoding TASK-3 in order to overexpress the channel specifically in the RGCs remaining after optic nerve injury. The effects of the two types of treatment were assessed based on the number and activity of RGCs, as well as overall image- and non–image-forming visual functions of the animals. Our results provide the first preclinical evidence that activating and upregulating TASK-3 may help protect and even restore visual function after optic nerve injury.

Materials and Methods

Mouse Model of Optic Nerve Injury

All animal procedures were performed according to the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research and approved by the Animal Care and Use Committee of Sichuan University (approval no. 2021201A). Male C57BL/6J mice, 6 to 8 weeks old (GemPharmatech, Nanjing, China), were housed at constant temperature and relative humidity on a 12-hour light/dark cycle with ad libitum food and water.

For experiments, animals were subjected to optic nerve crushing (ONC) to simulate injury caused by unilateral optic nerve compression,^9,10 which is the most common animal model of traumatic optic neuropathy.² Mice received tribromoethanol intraperitoneally (40 mg/mL), then a small slit was made using microsurgical scissors in the conjunctiva of the left eye at the 4 o'clock position. The peripheral tissue of the optic nerve and venous sinus were carefully separated using blunt tweezers until the optic nerve was clearly exposed. The optic nerve was crushed using tweezers for approximately 5 seconds at a position about 2 mm from the eyeball. Animals were allowed to recover on a heating pad and were then returned to their home cage. Sham animals were treated in the same way, except that the optic nerve was not crushed.

CHET3 Formulation for Eye Application

When delivered intraperitoneally, CHET3 acts mainly on the peripheral nervous system because it does not cross the blood–brain barrier efficiently,⁸ so we formulated it as a gel to apply to the eyes in the present study. β-Cyclodextrin powder (0.3g; Shanghai Yuanye Bio-Technology, Shanghai, China) was mixed well with normal saline (1 mL), then Poloxamer 407 gel powder (0.25g; BASF Pharma, Ludwigshafen, Germany) was added, and the resulting suspension was placed at 4°C until it became a transparent gel. Stock solution of CHET3 was suspended in this gel to a final concentration of 0.7 mg/mL for experiments. The stock solution was prepared by dissolving CHET3 (0.7 mg) in dimethyl sulfoxide (DMSO; 70 µL), followed by storage at −20°C.

Pharmacokinetics of CHET3 in the Mouse Eye

CHET3 gel at 0.7 mg/mL was applied once to one eye of wild-type mice, whose retinas were collected and homogenized at predefined time points (0.083, 0.25, 0.5, 0.75, 1.0, and 2.0 hours). CHET3 concentration in the retina was determined using liquid chromatography–tandem mass spectrometry as follows. Retinas were homogenized and centrifuged at 2000g for 10 minutes, and then supernatants were extracted with acetonitrile, which was injected onto a Shimadzu 30A high-performance liquid chromatography system (Shimadzu Scientific Instruments, Kyoto, Japan) equipped with an ACQUITY UPLC BEH C18 Column (2.1 × 100 mm, 1.7 mm; Waters, Milford, MA, USA) and coupled to a SCIEX Triple Quad 6500+ mass spectrometer (SCIEX, Framingham, MA, USA). The column temperature was 35°C, solvent A was 5-mM NH₄OAc with 0.1% formic acid in water, and solvent B was acetonitrile. Mean concentrations at each time point were determined from standard curves and were used to calculate pharmacokinetic area under the curve (AUC), C_max (maximum plasma concentration), T_max (time to reach maximum plasma concentration), and T_1/2 (time for concentration to reduce by half) in non-compartmental analyses in Phoenix WinNolin 6.4 (Certara, Radnor, PA, USA).

Patch-Clamp Electrophysiology of RGCs in Isolated Retinas

The firing activity of RGCs was assessed using patch-clamp electrophysiology as described elsewhere.⁷ Isolated retinas were incubated for at least 30 minutes in normal Ames’ saline (United States Biological Life Sciences, Salem, MA, USA) containing sodium bicarbonate (22.6 mM) through which 5% CO₂ and 95% oxygen had been bubbled for at least 30 minutes, after which the pH was adjusted to 7.4 using NaOH/HCl.

Each retina was incubated for 3 to 5 minutes with collagenase (1 mg/mL) and hyaluronidase (0.5 mg/mL) before use. The retinal piece was superfused with normal Ames’ saline at ∼2 mL/min. The activity of one RGC with a soma diameter of >15 µm was recorded per retinal piece in normal Ames’ saline containing 3-µM CHET3 or vehicle (DMSO).

For recording, pipettes were pulled from capillaries of fire-polished borosilicate glass (outer diameter, 1.5 mm; inner diameter, 1.1 mm; Sutter Instruments, Novato, CA, USA) using a Flaming/Brown P-1000 micropipette puller (Sutter Instruments). The resulting pipettes had an impedance of 4 to 8 MΩ and were filled with the following pipette solution during experiments: 125-mM potassium gluconate, 2-mM CaCl₂, 2-mM MgCl₂, 10-mM EGTA, 10-mM HEPES, 0.5-mM Mg–adenosine 5′-triphosphate, and bisodium–guanosine 5′-triphosphate (pH 7.2). Voltages recorded were corrected for a liquid junction potential of −13 mV, as estimated using Clampex (Molecular Devices, San Jose, CA, USA). During experiments, series resistance was monitored carefully, and data were discarded if resistance did not remain below 30 MΩ or 10% of the membrane resistance.

Immunostaining Against Key Proteins in Isolated Retinas

Retinas were transferred via glass pipette to a piece of nitrocellulose measuring 5 × 10 mm (0.8-µm pores, type AAWP; EMD Millipore, Burlington, MA, USA), and cut into a four-leaf clover shape. Each membrane with retina was fixed for 30 minutes using 4% paraformaldehyde and blocked for 60 minutes at room temperature in phosphate-buffered saline (PBS) containing 5% normal goat serum and 0.5% Triton X-100.

Retinas were incubated overnight with primary antibodies, incubated for 2 hours with secondary antibodies at room temperature, and mounted with Fluoromount-G (0100-20; SouthernBiotech, Birmingham, AL, USA). Primary antibodies recognized the RGC markers RNA-binding protein with multiple splicing (RBPMS; ab152101, 1:1000; Abcam, Cambridge, UK) and neuronal class β-tubulin III (TUJ1) (801202, 1:500; BioLegend, San Diego, CA, USA), or the intrinsic photosensitive RGC marker melanopsin (OPN4; ab19306, 1:1000; Abcam) and anti-POU4F2 rabbit pAB (BRN3B; P104486, 1:500; Epizyme Biotech, Shanghai, China). Secondary antibodies were Alexa Fluor 647–conjugated goat anti-mouse (111-605-003, 1:1000; Jackson ImmunoResearch Laboratories, West Grove, PA, USA) or Alexa Fluor 555– or Alexa Fluor 647–conjugated goat anti-rabbit (111-605-003, 1:1000; Jackson ImmunoResearch Laboratories).

In Situ RNA Hybridization

Eye balls were enucleated, fixed overnight at 4°C in 4% paraformaldehyde, dehydrated in 30% sucrose for 1 day at 4°C, and then sliced to a thickness of 15 µm on a cryostat.⁷ Kcnk9 mRNA encoding TASK-3 was detected through in situ hybridization using the RNAscope Multiplex Fluorescent Reagent Kit 2 (323100; Advanced Cell Diagnostics, Newark, CA, USA), RNAscope 4-Plex Ancillary Kit (323120; Advanced Cell Diagnostics), and specific probes (475681; Advanced Cell Diagnostics) according to the manufacturer's instructions.

For immunohistochemistry after in situ RNA hybridization, retinal sections were blocked in Tris-buffered saline (TBS) containing 10% normal goat serum (NGS) and 1% bovine serum albumin (BSA) at room temperature for 60 minutes. Slices were incubated overnight with rabbit antibodies against RBPMS (ab152101, 1:500; Abcam) at 4°C. Slices were washed in TBS–Tween 20 (three times, each lasting 5 minutes), and then incubated with Alexa Fluor 647 goat anti-rabbit secondary antibody (111-605-003, 1:500; Jackson ImmunoResearch Laboratories) at room temperature for 2 hours. To quantify Kcnk9-positive RGCs, Kcnk9 signal overlaps with the RBPMS signal were counted. Images were obtained on an A1R MP⁺ two-photon confocal scanning microscope (Nikon, Tokyo, Japan).

Visual Cliff Task

Visual function of mice was assessed in multiple ways. One binocular visual test was the visual cliff task, which assesses depth perception.^11,12 The “cliff” was a structure comprised of an upper transparent rectangular acrylic box measuring 84 × 53 × 50 cm and a lower one measuring 84 × 53 × 41 cm. All external sides of both boxes were covered with black paper to reduce reflection. The deep side of the lower box was covered with black-and-white checkerboard paper, where each square measured 3 × 3 cm. The shallow side of the upper box was also covered with the checkerboard paper, and the upper box was divided evenly into transparent and checkered parts to create the illusion of depth. Mice were free to move through both sides of the upper box. During the test, one mouse was placed in the middle of the upper box and its movements were recorded by camera for 5 minutes. The structure was cleaned with 30% alcohol and allowed to dry after each animal. Depth perception was semi-quantitated in terms of the discrimination index,¹³ defined as follows: (time spent at shallow end – time spent at deep end)/total time.

Visual Water Maze

Another test of binocular visual function was the visual water maze, which assesses visual acuity and contrast sensitivity.¹⁴ Two computer screens were set up facing one side of a Y-shaped tank.⁷ One screen showed a sine-wave grating pattern (0.12 cyc/deg), and the other showed a uniform gray pattern. Both patterns were generated using MATLAB (MathWorks, Natick, MA, USA) and Psychtoolbox. The patterns alternated randomly. Mice were released into the tank at the side farther from the screens and trained to swim to the side showing the grating pattern and then to escape from the water through a hidden platform. In each trial, the animal was considered to “pass” if it swam across the choice line and found the platform; otherwise, the animal was considered to “fail.” Mice were trained for two to four sessions a day, with each session comprised of five to 10 trials. Mice were considered trained when they achieved a pass rate above 80%, which typically occurred after 3 to 4 days (50–80 trials). Mice that did not learn the task by this time were excluded.

Visual acuity was assessed by displaying the grating initially at a spatial frequency of 0.12 cyc/deg, which increased stepwise by 0.06 cyc/deg each time that a mouse scored a pass. If the mouse failed, eight to 15 trials were performed at that frequency before advancing to the next one. The test was stopped when the pass rate fell below 70%, after which the pass rate was plotted against spatial frequency. The spatial frequency threshold was set at 70% accuracy.¹⁴ Contrast sensitivity was assessed in a similar manner, with the initial grating contrast of 0.12 cyc/deg (defined as 100%) decreasing by 10% each time.

Flash Visual Evoked Potentials

A third assessment of visual function was measurement of flash visual evoked potentials.^15,16 The amplitude of the biphasic potential wave reflects the integrity of the visual pathway from the injured eye to the contralateral cortex.^17–19 After deep anesthesia, the fur, skin, and skull membrane around the bregma were cut and carefully torn off. Small holes were drilled into the skull to a depth of about 400 µm at the contralateral prefrontal cortex (2.0 mm rostral to bregma) for placement of the reference electrode and at the contralateral primary visual cortex (3.6 mm caudal and 2.3 mm lateral to bregma) for placement of the recording electrode. Stainless-steel flathead screws, 0.8 × 4 mm, were placed into the holes to make slight contact with the cortex. Glass ionomer cement (Shanghai Medical Equipment, Shanghai, China) was used to fix the screws into position and seal the exposed area of the skull.

For experiments, animals were anesthetized as described above. One silver wire loop was touched to the reference electrode and another to the recording electrode, and the ground electrode was clamped onto the mouse tail. One eye was exposed to white flashes at a frequency of 1.0 Hz, bandpass of 0.5 to 85.0 Hz, and sampling frequency of 2000 Hz. While one eye was being tested, the other was covered with opaque black tape. Traces were recorded for 500 ms and stacked 100 times.

Pupillary Light Reflex

The non–image-forming activity of intrinsic photosensitive RGCs (ipRGCs) was assessed as described.⁷ Mice were dark adapted for at least 30 minutes, pupil size at baseline was recorded for 5 to 10 seconds, light was delivered for 5 to 10 seconds, and pupil size was measured during that period. Before experiments, light intensity was calibrated and measured using an optical power meter (8230; ADC, Saitama, Japan). Pupil size was analyzed using the “oval” tool in ImageJ (National Institutes of Health, Bethesda, MD, USA) at two time points: 1 second before light stimulation (baseline) and at 5 seconds after stimulation.

TASK-3 Overexpression in RGCs in Mice

TASK-3 was overexpressed specifically in RGCs by infecting mice with AAV carrying the Kcnk9 open reading frame (NCBI identifier 223604) fused at the 3′ end with sequences encoding two hemagglutinin tags and one FLAG tag. Expression of the transgene was driven by the Ple345(NEFL) promoter, which functions specifically in RGCs.²⁰ The original expression plasmid was pEMS2280 (111901; Addgene, Watertown, MA, USA), and the virus strain was AAV2/2-WY3239 (Taitool Bioscience, Shanghai, China). As a control, another recombinant virus was used that expressed enhanced green fluorescence protein bearing three FLAG peptides at the C-terminus (S0837-2-H50; Taitool Bioscience).

After deep anesthesia, the corneal limbus of one eye was punctured using a 30-gauge needle. Virus was then injected gently into the vitreous cavity of the eye whose optic nerve was crushed (2 µL, 0.5 × 10¹⁰ genomes per eye) using a 5-µL glass Hamilton syringe (Hamilton Company, Reno, NV, USA) equipped with a 33-gauge ultrafine removable needle (Hamilton Company). After 3 weeks, visual water maze experiments were performed, the animals were euthanized, and retinas were harvested for immunohistochemistry and in situ RNA hybridization as described above.

Statistical Analysis

The normality of continuous data was checked using the Shapiro–Wilk test, and normally distributed data were reported as mean ± SEM. Pairwise differences for variables showing a normal distribution were assessed for significance using a paired or unpaired two-tailed Student's t-test, and differences across more than two groups were assessed using two-way analysis of variance (ANOVA) followed by Sidak's post hoc correction for multiple comparisons. All statistical analyses were performed using Prism 8 (GraphPad Software, Boston, MA, USA). Results associated with P < 0.05 were considered statistically significant.

Results

TASK-3 Activation by CHET3 Renders RGCs Less Excitable

Given that the selective TASK-3 agonist CHET3 can reduce firing by dorsal root ganglion neurons and thereby chronic pain,⁸ we wondered whether it might reduce the excitability of RGCs, which also express TASK-3.⁷ Isolated retinas were light adapted and dissected under room lighting to minimize spontaneous activity. A current clamp was used to maintain their membrane potential near −70 mV. Stepped currents from 20 to 200 pA were then applied to elicit action potentials. At all current steps, CHET3 reduced the number of elicited action potentials (Figs. 1A, 1B). As a control, isolated retinas were exposed to DMSO, which has previously been shown not to affect elicited action potentials.⁷

Figure 1.

Selective TASK-3 agonist CHET3 decreases RGC excitability. (A) Examples of voltage traces of the RGCs in DMSO and CHET3 (3 µM) evoked by increasing current intensities. The current injection protocol is shown at the bottom. All RGCs were held at around −70 mV under a current clamp before current injections. (B) Number of spikes under different current injections in DMSO or CHET3 (n = 21 cells, paired t-test). (C) Panel C1 shows examples of initial action potentials of RGCs under DMSO or CHET3 (3 µM) conditions; C2, first derivatives of the action potentials in C1; and C3, phase plane plots of the action potentials in C1. (D–J) Bar graphs of the characteristics of action potentials under DMSO or CHET3 conditions: amplitude (D), half-width (E), maximum rise slope (F), maximum decay slope (G), threshold (H), RMP (I), and input resistance (J) (n = 25, paired t-test). (K) Average current–voltage relation (I–V) curve of RGCs in DMSO and in CHET3 (3 µM). (L) Bar graph of the current densities in DMSO and in CHET3 at +17 mV (n = 33 cells, paired t-test). (M) Graphic represents the subtraction trace from the representative CHET3-sensitive current in K. (N) Average I–V curve of the initiated inward current in DMSO and in CHET3 (3 µM, n = 18 cells). (O) Normalized conductance–voltage relationship curve in DMSO and in CHET3 (3 µM, n = 18 cells). (P) Bar graph of the half-inactivation voltage (V₅₀) in DMSO and in CHET3 (3 µM, n = 18 cells). All data are presented as mean ± SEM. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001; n.s., not significant.

Analysis of the profiles of initial elicited action potentials showed that CHET3 broadened action potentials (DMSO: 1.15 ± 0.08 ms; CHET3: 1.28 ± 0.09 ms) (Fig. 1E) and reduced their amplitude (DMSO: 71.1 ± 1.9 mV; CHET3: 64.1 ± 1.8 mV) (Fig. 1D), without significantly altering threshold, resting membrane potential (RMP), or input resistance (Figs. 1H, 1I). To isolate CHET3-sensitive potassium currents, we held RGCs at +17 mV for 500 ms, and then the holding voltage was ramped down from +17 to −133 mV within 500 ms (Supplementary Fig. S1). The currents were normalized by cell capacitance (mean ± SD, 71 ± 10 pF; n = 18 cells) and then plotted against the holding voltage. RGCs showed an outwardly rectifying current, which was significantly activated by 3-µM CHET3 (current density at +17 mV DMSO: 58.2 ± 3.9 pA/pF; CHET3: 65.9 ± 4.6 pA/pF) (Figs. 1K–M). We further tested the voltage-gated sodium channel activities by holding the voltage on different steps as shown in Supplementary Figures S1B and S1C. Although CHET3 did not alter the peak sodium current amplitude, it significantly increased the steady-state inactivation of these channels (V₅₀, DMSO: −50.6 ± 0.6 mV; CHET3: −52.7 ± 0.8 mV) (Figs. 1N–P). Notably, the kinetics of recovery from inactivation remained unaffected (Supplementary Figs. S1D, S1E).

Sustained TASK-3 Activation by CHET3 Protects RGCs From Death Induced by Optic Nerve Injury

ONC reduced the number of RGCs by 70% within 1 week, based on staining for the marker RBPMS (ONC: 200 ± 13/0.2 mm²; sham: 635 ± 34/0.2 mm²) (Figs. 2A–D). To examine whether CHET3 would alter this number, we formulated the compound within an eye gel. Pharmacokinetics showed that the maximum concentration (C_max) in retina was similar to the reported half-effective concentration (EC₅₀) to activate TASK-38 and was achieved at 0.08 hour after application to the corneal surface (Table). Therefore, we considered our corneal application to be sufficient to activate TASK-3 in retina. Sustained application of CHET3 for 1 week to the eyes of mice subjected to ONC led to numbers of RGCs that were approximately 5% higher than those in DMSO-treated animals (ONC-DMSO: 245 ± 7/0.2 mm²; ONC-CHET3: 220 ± 16/0.2 mm²) (Figs. 2E–H), whether in the periphery or center of the retina. CHET3 did not affect the number of RGCs in sham mice (Supplementary Fig. S2).

Figure 2.

Activation of TASK-3 protects RGCs from ONC injury. (A) Schematic of the experimental timeline for ONC surgery. (B, C) Confocal images of retinal wholemounts showing RBPMS-labeled RGCs after ONC (B) or sham surgery (C). Scale bar: 20 µm. (D) Bar graph of the number of RGCs after ONC or sham surgery (n = 5 or 6, unpaired t-test). (E) Schematic of the experimental timeline for CHET3 treatment after ONC surgery. (F, G) Confocal images of retinal wholemounts showing RBPMS-labeled RGCs under the treatment of DMSO (F) or CHET3 (G) after ONC. (H) Bar graph of the number of RGCs under the treatment of DMSO or CHET3 after ONC (n = 12, unpaired t-test). All data are presented as means ± SEM. *P < 0.05; ****P < 0.0001.

Table.

Pharmacokinetics of CHET3 in the Retina of Mice After a Single Administration (0.7 mg/mL) to the Ocular Surface

Parameter	Unit	Mean
T 1/2	Hour	1.15
Tmax	Hour	0.08
Cmax	ng/mL	436.99
AUC 0 – t	h·ng/mL	403.45
AUC 0 – inf	h·ng/mL	567.44

T_1/2, terminal half-life; T_max (median), time to C_max; C_max, maximum concentration; AUC, area under the concentrationtime curve.

Selective TASK-3 Agonist CHET3 Decreases RGC Excitability After ONC

To determine the intrinsic electrophysiological properties of large-soma RGCs and their response to CHET3 following injury, we performed recordings 7 days post-ONC. All identified large-soma RGCs exhibited a marked reduction in action potential frequency (Figs. 3A, 3B). Nevertheless, CHET3 continued to induce a significant decline in activity, further characterized by attenuated amplitudes (DMSO: 82.8 ± 1.9 mV; CHET3: 79.1 ± 2.0 mV) (Fig. 3D), broadened action potentials (DMSO: 0.88 ± 0.17 ms; CHET3: 0.86 ± 0.18 ms) (Fig. 3E), and slower firing kinetics (Figs. 3E–G). These changes were accompanied by a hyperpolarized threshold (DMSO: −44.3 ± 0.9 mV; CHET3: −45.6 ± 0.8 mV) (Fig. 3H) and decreased input resistance (DMSO: 0.24 ± 0.03 GΩ; CHET3: 0.22 ± 0.02 GΩ) (Fig. 3J). There was still no significant difference in RMP (DMSO: −62.9 ± 0.9 mV; CHET3: −63.2 ± 0.9 mV) (Fig. 3I). Notably, 15 out of 17 recorded RGCs (88%) maintained a CHET3-sensitive current post-injury, indicating the persistence of TASK-3 functional expression (Figs. 3K–M). We observed identical results of voltage-gated sodium channels in surviving RGCs 7 days post-ONC, as the steady-state inactivation was increased (V₅₀, DMSO: −55.3 ± 0.6 mV; CHET3: −56.9 ± 0.7 mV) (Figs. 3N–P), but the kinetics of recovery from inactivation was unchanged (Supplementary Figs. S1F, S1G).

Figure 3.

Selective TASK-3 agonist CHET3 decreases RGC excitability after ONC. (A) Examples of voltage traces of the RGCs from ONC eyes in DMSO and CHET3 (3 µM) evoked by increasing current intensities. The current injection protocol is shown at the bottom. All RGCs were held at around −70 mV under a current clamp before current injections. (B) Number of spikes under different current injections in DMSO or CHET3 (n = 19 cells, paired t-test). (C) Panel C1 shows examples of initial action potentials of RGCs from ONC eyes under DMSO or CHET3 conditions; C2, first derivatives of the action potentials in C1; and C3, phase plane plots of the action potentials in C1. (D–J) Bar graphs of the characteristics of action potentials under DMSO or CHET3 conditions: amplitude (D), half-width (E), maximum rise slope (F), maximum decay slope (G), threshold (H), RMP (I), and input resistance (J). (K) Average current–voltage relation (I–V) curve of RGCs from ONC eyes in DMSO and in CHET3 (3 µM). (L) Bar graph of the current densities in DMSO and in CHET3 at +17 mV. (M) Graphic represents the subtraction trace from the representative CHET3-sensitive current in K. (N) Average I–V curve of the initiated inward current in DMSO and in CHET3 (3 µM). (O) Normalized conductance–voltage relationship curve in DMSO and in CHET3 (3 µM). (P) Bar graph of the half-inactivation voltage (V₅₀) in DMSO and in CHET3 (3 µM). All data are presented as mean ± SEM. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001.

Daily TASK-3 Activation by CHET3 After Optic Nerve Injury Retains Image- and Non–Image-Forming Visual Function

We sought to explore whether the observed ability of CHET3 to protect RGCs from death after optic nerve injury would translate to greater visual function. Indeed, analysis of flash visual evoked potentials indicated that CHET3 led to stronger amplitudes than DMSO vehicle (sham: 50.5 ± 9.4 µV; ONC-DMSO: 25.0 ± 2.2 µV; ONC-CHET3: 38.4 ± 3.6 µV) (Figs. 4A, 4B), suggesting that activating TASK-3 can protect the monocular visual pathway from the retina to the primary visual cortex. Similarly, CHET3 led to better depth perception in mice, reflected in greater ability to perceive the visual cliff, leading the animals to spend more time on the shallow side of the test chamber (sham: 0.67 ± 0.05; ONC-DMSO: 0.12 ± 0.08; ONC-CHET3: 0.39 ± 0.07) (Figs. 4C, 4D). CHET3 rescued most of the visual acuity loss induced by ONC (sham:0.53 ± 0.03 cyc/deg; ONC-DMSO: 0.29 ± 0.03 cyc/deg; ONC-CHET3: 0.41 ± 0.03 cyc/deg), and it led to a contrast sensitivity threshold that was 21 percentage points lower than that in DMSO controls and close to that in sham animals (sham: 0.16 ± 0.01; ONC-DMSO: 0.43 ± 0.03; ONC-CHET3: 0.23 ± 0.02) (Figs. 4F, 4G). These results suggest that CHET3 rescues the loss of image-forming visual activity mediated by RGCs.

Figure 4.

TASK-3 activation preserved image-forming visual function. (A) Representative responses of flash visual evoked potentials (f-VEP) recordings from sham mice or DMSO- or CHET3-treated eyes in ONC mice. (B) Quantification of f-VEP amplitudes (n = 6 to 17, one-way ANOVA). (C) Schematic of the visual cliff test. (D) Bar graph of depth perception in each group (n = 6 to 12, one-way ANOVA). (E) Schematic of visual water maze. (F, G) Bar graph for spatial frequency (n = 6 to 16, one-way ANOVA) (F) and contrast sensitivity (n = 6 to 19, one-way ANOVA) (G). All data are presented as mean ± SEM. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001.

Next we examined whether CHET3 could retain loss of non–image-forming visual activity mediated by ipRGCs, which are atypical, melanopsin-expressing photoreceptor cells,^21,22 more than 90% of which express TASK-3.⁷ These cells contribute to non–image-forming functions, such as the pupillary light reflex (PLR). CHET3 helped protect ipRGCs from death induced by optic nerve injury (sham: 21.0 ± 1.6/0.4 mm²; ONC-DMSO: 12.3 ± 0.7/0.4 mm²; ONC-CHET3: 17.9 ± 1.4/0.4 mm²) (Figs. 5A–D), consistent with its ability to protect other types of RGCs. CHET3 also partially retained the ability of the injured pupil to constrict after exposure to intense light (13 log quanta/cm²/s; sham: 10.9% ± 0.5%; ONC-DMSO: 76.5% ± 4.4%; ONC-CHET3: 54.7% ± 5.5%) (Figs. 5E, 5F). CHET3 had no effect on the visual function in sham mice (Supplementary Fig. S2).

Figure 5.

TASK-3 activation preserved non–image-forming visual function. (A–C) Confocal images of retinal wholemounts showing immunostaining of OPN4 (magenta) and TUJ1-labled (green) RGCs from sham mice (A), and DMSO-treated (B) or CHET3-treated (C) eyes in ONC mice. Scale bar: 50 µm. Arrows show OPN4-positive cells. The zoom-in view of the white box is on the right. Scale bar: 20 µm. (D) Bar graph of ipRGC numbers from sham mice or DMSO-treated or CHET3-treated eyes in ONC mice (n = 6 to 25, one-way ANOVA). (E) Example images of PLR from sham mice and DMSO-treated or CHET3-treated eyes in ONC mice under different light conditions. (F) Irradiance response relationship (IRR) of PLR in sham mice and DMSO-treated or CHET3-treated eyes in ONC mice. (n = 6 to 11, two-way ANOVA). All data are presented as mean ± SEM. *P < 0.05; **P < 0.01.

TASK-3 Overexpression in the Residual RGCs 1 Week After Optic Nerve Injury Restores Image- and Non–Image-Forming Visual Functions

The above experiments indicated that daily pharmacological activation of TASK-3 can protect RGCs and their image- and non–image-forming functions at 1 week after optic nerve injury. Next, we wanted to know whether overexpression of TASK-3 in the approximately 30% of RGCs remaining at 1 week after optic nerve injury could exert similar therapeutic effects. We thought this would be likely given our finding that TASK-3 overexpression can reverse age-related visual decline in mice.⁷

We generated AAV2-Ple345(NEFL)-kcnk9-HA (AAV2–TASK-3) and AAV2-Ple345(NEFL)-EGFP (AAV2-control). Animals were subjected to ONC, then injected intravitreally 1 week later with AAV encoding TASK-3 or control virus. At 4 weeks after injury, the animals were assessed in the visual water maze, then their retinal tissue was harvested for immunostaining against key proteins and analysis of TASK-3 mRNA (Fig. 6A). We identified RGCs based on their expression of the marker protein TUJ1. By 4 weeks after injury, more than 90% of the original population of RGCs had died, yet the numbers of residual cells were larger in animals overexpressing TASK-3 (sham: 551.7 ± 25.7/0.4 mm²; ONC-AAV2-control: 48.7 ± 4.9/0.4 mm²; ONC-AAV2-TASK-3: 66.7 ± 6.2/0.4 mm²) (Figs. 6B–E). As expected, the AAV was expressed specifically in RGCs and not in other retinal layers (Fig. 6F). After ONC, the expression level of TASK-3 mRNA was down with the decrease in the number of RGCs. Overexpression of TASK-3 specifically in the remaining RGCs rescued the level of TASK-3 mRNA similar to that in sham group (sham-AAV2-control: 14.0 ± 1.8 µm²/cell; ONC-AAV2-control: 2.1 ± 0.3 µm²/cell; ONC-AAV2-TASK-3: 15.3 ± 3.5 µm²/cell) (Figs. 6F–I). Although the rescued cells were relatively few, their overexpression of TASK-3 was sufficient to restore visual acuity (sham-AAV2-control: 0.54 ± 0.05 cyc/deg; ONC-AAV2-control: 0.30 ± 0.03 cyc/deg; ONC-AAV2-TASK-3: 0.41 ± 0.03 cyc/deg) and contrast sensitivity (sham-AAV2-control: 0.17 ± 0.02; ONC-AAV2-control: 0.42 ± 0.03; ONC-AAV2-TASK-3: 0.28 ± 0.02) (Figs. 6J, 6K), nearly to the level in the sham group.

Figure 6.

Overexpressing Kcnk9 specifically in RGCs restores the image-forming visual function in ONC mice. (A) Schematic of the experimental timeline for TASK-3 overexpression after ONC surgery. (B–D) Confocal images of retinal wholemounts showing TUJ1-labeled RGCs from the eye injected with control virus in sham mice (B), the eye injected with control virus in ONC mice (C), and the eye injected with TASK-3 overexpressing virus in ONC mice (D). Scale bar: 50 µm. (E) Bar graph of ipRGCs number from control virus–injected sham mice, control virus–injected ONC mice, and TASK-3 overexpressing virus–injected ONC mice (n = 6 to 9; unpaired t-test). (F) RNAscope in situ hybridization for TASK-3 (Kcnk9, magenta) and immunostaining for RBPMS (green) in sham mouse retina injected with the control virus AAV2-Ple345(NEFL)-EGFP (left). Zoom-in view of the white box is on the right. ONL, outer nuclear layer; INL, outer nuclear layer; DAPI (blue), 4′,6-diamidino-2-phenylindole. Scale bar: 50 µm. (G, H) RNAscope in situ hybridization for TASK-3 (Kcnk9, magenta) and immunostaining for RBPMS (green) in ONC mouse retina injected with control virus (G) and TASK-3 (Kcnk9) overexpressing virus (H). (I) Bar graph of TASK-3 (Kcnk9) RNAscope signal area normalized by cell numbers at the ganglion cell layer (n = 5 to 11, one-way ANOVA). (J, K) Bar graph for spatial frequency (n = 6 to 26, one-way ANOVA) (J) and contrast sensitivity threshold (n = 6 to 26, one-way ANOVA) (K). All data are presented as mean ± SEM. *P < 0.05; **P < 0.01; ***P < 0.001.

Similarly, TASK-3 overexpression improved survival of ipRGCs (sham-AAV2-control: 22.5 ± 1.9/0.4 mm²; ONC-AAV2-control: 9.1 ± 1.3/0.4 mm²; ONC-AAV2-TASK-3: 14.3 ± 2.1/0.4 mm²) (Figs. 7A–D), and their PLR under low light intensity condition at 4 weeks after injury (10 log quanta/cm²/s; sham-AAV2-control: 60.4% ± 5.1%; ONC-AAV2-control: 78.5% ± 3.7%; ONC-AAV2-TASK-3: 63.6% ± 4.8%) (Figs. 7E, 7F). To determine if specific ipRGC subtypes are preferentially retained, we performed dual-labeling for OPN4 and BRN3B to differentiate between M1 and non-M1 populations.²³ Consistent with the established molecular profile where classical M1 ipRGCs are characterized as OPN4⁺/BRN3B^–,^24,25 we observed that these cells exhibited higher survival rates post-ONC compared to their BRN3B-expressing counterparts (OPN4⁺/BRN3B⁺ and OPN4^low/BRN3B⁺). Notably, AAV-mediated TASK-3 overexpression significantly increased the survival of ipRGC. Furthermore, the relative proportion of M1-type ipRGCs in TASK-3–treated retinas was shifted back to levels observed in sham controls (Figs. 7G–J), suggesting that TASK-3 acts as a potent survival factor across the entire ipRGC population.

Figure 7.

Overexpressing Kcnk9 specifically in RGCs restores the non–image-forming visual function in ONC mice. (A–C) Confocal images of retinal wholemounts showing immunostaining of OPN4 (magenta) and TUJ1-labled (green) RGCs from the eye injected with control virus in sham mice (A), the eye injected with the control virus in ONC mice (B), and the eye injected with TASK-3 overexpressing virus in ONC mice (C). Arrows show OPN4-positive cells. Scale bar: 50 µm. The zoom-in view of the white box is on the right. Scale bar: 20 µm. (D) Bar graph of ipRGC numbers from control virus–injected sham mice, control virus–injected ONC mice, and TASK-3 overexpressing virus–injected ONC mice (n = 6 to 9, one-way ANOVA). (E) Example images of PLR from the eye injected with control virus in sham mice, the eye injected with control virus in ONC mice, and the eye injected with TASK-3 overexpressing virus in ONC mice under different light conditions. (F) Irradiance response relationship (IRR) of PLR from the eye injected with control virus in sham mice, the eye injected with control virus in ONC mice, and the eye injected with TASK-3 overexpressing virus in ONC mice (n = 6 to 17, two-way ANOVA). (G–I) Immunostaining of BRN3B (green) and OPN4 (magenta) in sham mouse retina injected with control virus (G) and in ONC mouse retina injected with control virus (H) or TASK-3 (Kcnk9) overexpressing virus (I). Scale bar: 20 µm. (J) Percentage of ipRGC subtypes in sham mouse retina injected with control virus and in ONC mouse retina injected with control virus or TASK-3 (Kcnk9) overexpressing virus (n = 6, one-way ANOVA). All data are presented as mean ± SEM. *P < 0.05; ***P < 0.001.

Discussion

Our experiments provide the first evidence that TASK-3 can substantially influence visual function after optic nerve injury and that activating or overexpressing it can protect RGCs from injury-induced death while also retaining or restoring their activity underlying image formation and other activities such as the PLR. Two-pore domain potassium channels such as TASK-3 are believed to maintain RMP and regulate neuronal excitability.²⁶ Our previous work established that inhibition or deletion of TASK-3 from RGCs increases their excitability by increasing input resistance and depolarizing the RMP.⁷ In the present study, we showed that activating TASK-3 in RGCs decreases their excitability by slowing action potentials, without affecting input resistance, RMP, or the action potential threshold. The lack of a significant effect on RMP may be attributable to the heterogeneity of RGC subtypes. The RGCs with a soma diameter of >15 µm were recorded in our system, which could be ON-transient, ON-sustained, OFF-transient, and OFF-sustained alpha RGCs.^27,28 The recorded population likely includes multiple RGC subtypes, which may contribute to the variability observed in our TASK-3 activation data. To minimize some light-derived activity from upstream circuities, mice and retina were light adapted before and during experiments. Although light adaptation reduced upstream signaling, intrinsic biophysical properties could account for the heterogeneity in RMP. TASK-3, a subtype of the two-pore domain potassium channel family, exhibits outwardly rectifying properties. The lack of effect on RMP following TASK-3 activation by CHET3 is also likely due to the minimal conductance of these channels at voltages between −60 and −80 mV. In dorsal root ganglia, neither TASK-3 activation with CHET3 nor selective TASK-3 inhibition with PK-THPP affected their RMP.⁸ These findings suggest that the influence of TASK-3 on RMP depends on the type of neuron and the situation, which may reflect that different compensatory mechanisms or secondary signaling cascades kick in when the potassium current changes. Further investigations are warranted to elucidate the functional heterogeneity observed across different cells, and future studies employing dark-adapted recordings will be essential to determine how TASK-3 modulation affects RGC circuit-level signaling and specific visual pathways.

Beyond its modulation of TASK-3 channels, our findings reveal that CHET3 influences voltage-gated sodium channel kinetics, specifically by enhancing steady-state inactivation without altering recovery rates. This shift reduces the pool of available sodium channels during repetitive firing, thereby limiting metabolic demand and preventing excitotoxic injury. Such a mechanism is highly consistent with established neuroprotective strategies wherein the stabilization of inactivated sodium channels prevents neuronal overexcitation. In the context of the injured retina, the dual action of CHET3—increasing TASK-3–mediated leak conductance while simultaneously promoting voltage-gated sodium channel inactivation—likely provides a synergistic effect that stabilizes the activity and preserves the morphological and physiological integrity of resilient RGC subtypes such as alpha RGCs.

Our data demonstrate that TASK-3 mRNA levels decline following ONC, likely due to a combination of transcriptional downregulation within individual neurons and the overall loss of RGCs. Notably, the majority of recorded large-soma RGCs, a population known for its selective survival post-ONC injury, retained a functional TASK-3–mediated current. This suggests that RGCs with higher TASK-3 expression may be less susceptible to injury-induced degeneration. These findings align with the subtype-specific resilience described in previous studies^29,30 and further suggest that elevated TASK-3 expression may confer a neuroprotective advantage against degeneration.

We attribute the protective and restorative effects of TASK-3 activation or overexpression to its ability to mitigate excitotoxicity following optic nerve injury.^31,32 Excessive glutamate binds to ionotropic glutamate receptors, triggering an influx of Ca²⁺ and subsequent organelle stress and apoptosis.^33–35 TASK-3 activation by CHET3 slowed the generation and propagation of single-action potentials, which may mitigate overactivation of glutamate receptors and the resulting excitotoxicity. Recent work has shown that activating Ca²⁺/calmodulin-dependent protein kinase II can promote RGC axon regeneration after injury.³⁶ Previous studies have demonstrated that depolarization-induced Ca²⁺ influx dynamically regulates Kcnk9 (TASK-3) transcription through downstream effectors such as calcineurin, thereby modulating intrinsic excitability.^37,38 Consequently, activity-dependent Ca²⁺ signaling may facilitate RGC axon regeneration by fine-tuning TASK-3 expression, ensuring the neuronal excitability levels required for sustained regenerative growth, so future studies should examine whether this kinase helps mediate the therapeutic effects of TASK-3 activation or overexpression in these cells.

Our results suggest relatively strong therapeutic benefits of TASK-3 activation or overexpression, given that mono- and binocular vision function was substantially restored despite the loss of large numbers of RGCs in the mouse model. It is possible that the benefits might have been even greater at higher doses. The maximal concentration of CHET3 in retina after a single administration to the ocular surface was close to the EC₅₀ of the drug, which meant that at least half of TASK-3 channels should have been activated. On the other hand, the terminal half-life of CHET3 in retina in our system was only 1.15 hours. Therefore, higher or more frequent dosing may be needed to optimize benefit. This should be explored in preclinical studies of our CHET3 formulation as an eye gel.

TASK-3 activation or overexpression benefited not only RGCs as a whole but also the specific subset of ipRGCs, which transmit image-forming signals like other RGCs, as well as non–image-forming signals, which regulate circadian rhythms, emotions, and the PLR.^39–41 At scotopic or mesopic light intensities, the PLR is driven mainly by the photoreceptor-mediated ipRGCs signal; however, at high light intensity, the reflex is directly dominated by melanopsin-mediated signal from ipRGCs.^42–44 The numbers of ipRGCs in our experiments involving corneal CHET3 application were higher than those in our experiments involving TASK-3 overexpression. Thus, CHET3 may restore sensitivity to the PLR at high light intensity, whereas TASK-3 overexpression may restore the sensitivity of the reflex at scotopic or mesopic light intensities. In other words, pharmacological activation and overexpression of TASK-3 may operate via different mechanisms to affect the PLR, which requires further investigation.

Optic nerve injury usually leads to irreversible vision loss in the clinic, despite surgical and drug treatments. Various studies have focused on protecting RGCs and promoting the regeneration of their axons.^2,36 Here, we provide evidence that TASK-3 overexpression in residual RGCs, even when they number only about 10% of the population after optic nerve injury, can restore visual acuity and the contrast sensitivity threshold nearly completely. Our findings justify further exploration of whether activating or upregulating TASK-3 is a promising therapy to improve the prognosis of patients who have suffered optic nerve injury. The same approaches should also be explored for patients with glaucoma or optic glioma, both of which induce loss of RGCs.² These approaches may have even wider relevance, given that TASK-3 activation has been shown to provide good peripheral analgesia.^8,45

Although our study has highlighted the role of TASK-3 channels in RGCs, our RNAscope analysis indicate that Kcnk9 mRNA is also present within the inner nuclear layer (INL) and, to a lesser extent, in Müller glia. This distribution suggests that the neuroprotective efficacy of CHET3 may stem from a multilocus action within the retinal circuitry. Specifically, the activation of TASK-3 in Müller cells could enhance potassium conductance and metabolic support, whereas its presence in the INL may modulate interneuron-mediated inhibition. However, given the robust expression of TASK-3 in RGCs and the direct stabilization of their intrinsic excitability by CHET3, we propose that RGC TASK-3 activation remains a critical determinant of neuronal survival following axonal injury. This broader expression profile suggests that CHET3 may offer a synergistic, panretinal protective effect, reinforcing the therapeutic potential of TASK-3 activation in treating optic neuropathies.

In conclusion, our study identified TASK-3 (Kcnk9) as a pivotal determinant of RGC survival following axonal injury. We demonstrated that, whereas ONC triggers a significant downregulation of TASK-3 mRNA, large-soma RGCs and M1-type ipRGCs maintain a functional TASK-3–mediated leak conductance that correlates with their inherent resilience. Mechanistically, we showed that the small molecule activator CHET3 provides neuroprotection through a synergistic dual action: enhancing TASK-3–mediated membrane stabilization and modulating voltage-gated sodium channel kinetics by promoting steady-state inactivation. This combined effect effectively prevents excitotoxic failure in injured neurons. Furthermore, viral-mediated overexpression of TASK-3 proved sufficient to rescue both classical RGCs and ipRGC subpopulations, restoring the phenotypic distribution of the retina to near-sham levels. Collectively, these findings validate TASK-3 as a viable therapeutic target for preserving visual circuitry and offer a promising pharmacological strategy for the treatment of traumatic optic neuropathies and other diseases, such as glaucoma.