Abstract
Increasing spike rates drive greater neuronal energy demand. In turn, mitochondrial ATP production leads to the generation of reactive oxygen species (ROS) that can modulate ion channel gating. Does ROS production autoregulate the excitability of a neuron? We investigated the links between retinal ganglion cell (RGC) excitability and spike activity-driven ROS production in male and female mice. Changes to the light-evoked and current-evoked spike patterns of functionally identified αRGC subtypes, along with their NaV channel-gating properties, were recorded during experimentally induced decreases and increases of intracellular ROS. During periods of highest spike rates (e.g., following light onset in ON sustained RGCs and light offset in OFF sustained RGCs), these αRGC subtypes responded to reductions of ROS (induced by catalase or glutathione monoethyl ester) with higher spike rates. Increases in ROS (induced by mercaptosuccinate, antimycin-A, or H2O2) lowered spike rates. In ON and OFF transient RGCs, there were no changes in spike rate during ROS decreases but increased ROS increased spiking. This suggests that endogenous ROS are intrinsic neuromodulators in RGCs having high metabolic demands but not in RGCs with lower energy needs. We identified ROS-induced shifts in the voltage-dependent gating of specific isoforms of NaV channels that account for the modulation of ON and OFF sustained RGC spike frequency by ROS-mediated feedback. ROS-induced changes to NaV channel gating, affecting activation and inactivation kinetics, are consistent with the differing spike pattern alterations observed in RGC subtypes. Cell-autonomous generation of ROS during spiking contributes to tuning the spike patterns of RGCs.
SIGNIFICANCE STATEMENT Energy production within retinal ganglion cells (RGCs) is accompanied by metabolic by-products harmful to cellular function. How these by-products modulate the excitability of RGCs bears heavily on visual function and the etiology of optic neuropathies. A novel hypothesis of how RGC metabolism can produce automodulation of electrical signaling was tested by identifying the characteristics and biophysical origins of changes to the excitability of RGCs caused by oxidizing by-products in the retina. This impacts our understanding of the pathophysiology of RGC dysfunction, supporting an emerging model in which increases in oxidizing chemical species during energy production, but not necessarily bioenergetic failure, lead to preferential degeneration of specific subtypes of RGCs, yielding loss of different aspects of visual capacity.
Introduction
To convey visual information to the brain, retinal ganglion cells (RGCs) generate action potentials that traverse long sections of unmyelinated axon across the retinal surface before entering the optic nerve. Maintaining the transmembrane ionic gradients required for signaling with action potentials in unmyelinated nerves requires great expenditure of ATP. Inherited and environmental optic neuropathies often involve mitochondrial proteins and despite their ubiquitous expression in all neurons, RGCs are preferentially lost, underscoring the constraints imposed by metabolism on RGC function. RGC subtypes have unique biophysical properties (O'Brien et al., 2002; Margolis and Detwiler, 2007; Wong et al., 2012), which vary considerably between the subtypes (Emanuel et al., 2017; Werginz et al., 2020). While the ion channel properties of RGCs have been investigated using patch-clamp analysis of isolated cells (Lukasiewicz and Werblin, 1988), no reports describe the ion channel complement and voltage-dependent kinetics that determine the unique electrogenic phenotypes of RGC subtypes in situ. A theoretical framework, metabolically efficient coding (Crotty et al., 2006), seeks to explain the evolutionary constraints that shape the diverse electrogenic phenotypes of different neuronal subtypes, proposing that keeping metabolic expense under control constrains the biophysical properties of neurons in the trade-off between bandwidth (spike rate) and energetic cost (Carter and Bean, 2009; Hasenstaub et al., 2010). “Brisk” αRGCs provide high-bandwidth channels, whereas most RGCs spike at low frequencies (Koch et al., 2004, 2006), below maximal information transfer rates.
Metabolic by-products, including reactive oxygen species (ROS) produced when mitochondria generate ATP, are potent neuromodulators (Lee et al., 2015), altering biophysical factors including neurotransmitter-gated and voltage-gated channels. Prolonged periods of increased spiking raise ROS levels (Avshalumov et al., 2005). Oxidants target multiple proteins involved in neuronal excitability including GABA receptors (Accardi et al., 2014), TRP channels (Chuang and Lin, 2009), voltage-gated NaV (Evans and Bielefeldt, 2000; Kassmann et al., 2008), CaV (Li et al., 1998), and KV channels (Dantzler et al., 2019), and ATP-gated K+ channels (Avshalumov et al., 2005). Even within ion channel classes, such as the NaV classes, oxidant effects differ considerably between molecular isoforms (Kassmann et al., 2008; Schlüter and Leffler, 2016), complicating predictions of how ROS affect neuronal function.
RGCs are broadly categorized by the kinetics (sustained vs transient) and polarity (ON vs OFF) of their light responses. Sustained subtypes produce tonic spiking during light responses while transient cells spike phasically in response to light onset or offset. Analysis of mouse RGCs shows ∼17 sustained and ∼16 transient subtypes (Goetz et al., 2022). Sustained αRGCs have high basal activity levels and thus high metabolic demand, while transient cells produce sparse responses with lower metabolic demand. In mice, αRGCs comprise four cell types divided into ON and OFF sustained and ON and OFF transient subtypes (Krieger et al., 2017), which produce high spike rates and have fast axonal conduction velocities.
The investigations reported here tested the hypothesis that the elevated spike rates of sustained αRGCs, but not transient αRGCs, increase metabolic demand producing greater ROS production, and that increased ROS levels then act as feedback signals to slow spiking. Light stimulation of identified αRGC subtypes with experimentally enhanced intracellular ROS scavenging or artificially increased ROS, permitted the characterization of changes in light-evoked spike patterns. The results suggest that ROS act intrinsically to modulate responses to light in RGCs during tonically elevated metabolic demand. The data also show that NaV1.1/1.3 channels, which facilitate sustained spiking in response to light, are more susceptible to modulation by ROS leading to reduced spiking. This evidence suggests that NaV channel subtypes responsible for continuous, rapid spiking in sustained cells are the ones most sensitive to ROS, enabling both lasting high spike rates and an automatic reduction of spike rate when ROS levels become excessive. This may reflect an evolutionary solution that actively guards the most energy stressed RGCs against high levels of ROS, triggered when the RGCs produce elevated ATP and ROS during periods of high energy demand.
Materials and Methods
Patch-clamp recording of dark-adapted RGCs in isolated mouse retina.
Experiments were performed using genetically labeled mouse αRGCs in accordance with the ARVO (The Association for Research in Vision and Ophthalmology) “Statement for the Use of Animals in Ophthalmic and Vision Research” and with the guidelines for the welfare of experimental animals issued by the US Public Health Service “Policy on Human Care and Use of Laboratory Animals” (2002). All recordings of light-induced and current-induced responses, as well as voltage-clamp analysis of NaV channel currents, were made using the Kcng4-cre::tdTomato mouse line (Krieger et al., 2017). A subset of experiments tested the light responsivity of type 1 melanopsin-containing, intrinsically photosensitive RGCs from the OPN4EGFP mouse line (MRRC#033064-UCD; Matynia et al., 2016). Male and female mice were deeply anesthetized with 1–3% isoflurane (IsoFlo, Abbott Laboratories) and decapitated. The eyes were then enucleated in dim light, and the anterior portion including the lens was removed. The resulting eyecup was trimmed, and a section of retina placed vitreal side up in a superfusion chamber was mounted on the microscope. These flat-mounts typically included a layer of full or partial pigmented epithelial cells under the photoreceptor layer. Retinas were superfused via a gravity driven fast flow system (ALA) with a bath solution containing the following (in mm): 120 NaCl, 2 CaCl2, 3 KCl, 1 MgCl2, 1.2 NaH2PO4, 10 glucose, and 25 NaHCO3, bubbled continuously with 95% O2/5% CO2 and maintained at 34°C with an inline feedback-regulated heater (ALA). This was preferred over other commonly used solutions, such as the Ames test, that contain ingredients having significant redox-regulating properties.
Patch electrodes with 2–8 MΩ open tip resistance were pulled from fire-polished borosilicate glass capillary tubes using a micropipette puller (Sutter Instrument). The bath reference electrode consisted of an AgCl wire in a side chamber. Cell voltage was clamped with an amplifier (MultiClamp 700B, Molecular Devices) using whole-cell capacitance and series resistance compensation. Current signals were filtered at 5 kHz and digitized at 10 kHz with a digitizer (Digidata 1440A, Molecular Devices) for storage on the hard disk of a computer running pCLAMP 9 acquisition software (Molecular Devices). Patch pipettes were filled with the following (in mm): 10 KCl, 120 K-gluconate, 0.5 CaCl2, 5 EGTA, 10 HEPES, 4 Mg-ATP, 0.4 lithium-GTP, and 10 phosphocreatine, and buffered to pH 7.2. Recordings using the Kcng4-cre::tdTomato line fit the established αRGC electrophysiological statistics and structural features of large somata and stout axons (Peichl, 1991; Krieger et al., 2017). The tdTomato-expressing type 5 bipolar and displaced amacrine cells were easily identified, and recording from them was avoided. Difficulties involving space clamp (e.g., voltage clamp of all cellular compartments, including the long RGC axon) hinder the study of large and fast voltage-gated currents in intact RGCs, and since acutely isolated RGCs, where space clamp is optimized, cannot be identified by their light response, we used the outside-out nucleated patch technique in identified, light-responsive RGCs in the ex vivo flat-mount retinal preparation (Lipton and Tauck, 1987). Nucleated patch recordings used a pipette solution containing the following (in mm): 120 CsMeSO4, 2.8 NaCl, 0.3 CaCl2, 5 EGTA, 20 TEACl, 2 Mg-ATP, 0.5 Na3-GTP, 10 HEPES, and 10 Na2-phosphocreatine, at pH 7.3. Nucleated patches were corrected for a measured liquid junction potential of −7.5 mV. Series resistance was compensated by 70–100%.
Patch-clamp recordings were made using infrared light for cell visualization to maintain dark-adapted conditions, and td-Tomato-expressing RGCs were identified with one to two brief fluorescence-evoking green light steps (∼500 ms). Once a cell was targeted, a dark waiting period of 5–10 min was observed before patching the cell under infrared illumination to minimize any light adaptation effects. Green light steps (530 ± 10 nm; SOLA TRITC emission, Nikon) were then applied to determine whether the αRGC was sustained or transient and ON or OFF type. Light steps had a stimulus intensity at 530 nm of 5.12 × 106 photons/μm2/s corresponding to 2.26 × 106 R*/rod/s assuming a light-collecting area for mouse rods of 0.63 μm2 (Smeds et al., 2019) and using the mouse rhodopsin absorbance spectrum (Wang et al., 2011). This stimulus corresponded to 7.37 × 105 isomerization/m cone opsin/cone/s, assuming a collecting area of 0.12 μm2 (Sakurai et al., 2011). Testing the effects of endogenous ROS level changes on RGC excitability was done with current-clamp recordings of spontaneous, light-evoked, and current-evoked spiking while (1) reducing normal levels of ROS by adding catalase (2000–5,000 U/mg protein; diluted to 200–500 IU/ml; catalog #C1345, Sigma-Aldrich) after solubilization, to the pipette solution; (2) increasing ROS level by adding antimycin-A (2 μm; catalog #A8674, Sigma-Aldrich) to the pipette solution; or (3) reversibly increasing ROS levels in the entire retina using mercaptosuccinate (MCS; 1 mm; catalog #88460, Sigma-Aldrich) in the superfusion solution or with nonreversible puff applications of glutathione monoethyl ester (GME; catalog #G1404, Sigma-Aldrich) or H2O2 (100 μm; catalog #H1009, Sigma-Aldrich) near the cell soma (Avshalumov et al., 2005). ROS increases during voltage-clamp recordings of NaV channels used short puffs of H2O2. Sodium channels were blocked with tetrodotoxin (TTX; 1 μm; catalog #T8024, Sigma-Aldrich) or the NaV1.1/1.3 subtype-specific blocker ICA121431 (500 nm; catalog #5066, Tocris Bioscience).
Experimental design and statistical analysis.
All data are reported as the mean ± SEM. Data were analyzed using Clampfit 10.3 (Molecular Devices). Graphing and statistical analyses were performed using MATLAB (MathWorks), Python 3.7, R (University of Vienna), and/or Excel (Microsoft). Statistical significance was tested with paired t tests, or when more than two treatments were involved, ANOVA with post hoc Tukey's HSD test. Spike properties (see Figs. 6, 10) were tested for significance using nonparametric tests including the Mann–Whitney U test or the Kruskal–Wallis test with post hoc Dunn's test if more than two treatments were involved. This was because of significant bimodality in the experimental outcomes. The statistics program R only reported up to approximately p < 10−10 with high n values; hence, this value is shown in these cases, otherwise we denote p values as the actual computed value with values of <0.05 considered to be statistically significant and are denoted in the figures with one asterisk, while p < 0.01 and p < 0.001 are denoted with two and three asterisks, respectively.
Results
Changes to spiking in RGCs occur in response to elevated natural oxidizing species levels
Light-evoked spiking drives rapid metabolic demand and reveals that different RGC subtypes consume energy at very different rates (Perge et al., 2009). We investigated whether metabolic by-products might act intrinsically to modulate RGC spiking and whether such effects might depend on the basal metabolic demand on the cell. Kcng4-cre::tdTomato mice (Krieger et al., 2017) were used since these mice have four types of fluorescently labeled RGCs that can be divided by the average spike rate, and thus metabolic demand, both during spontaneous and light-evoked activity. ON and OFF sustained cells have relatively high levels of spontaneous activity and spike for the entire duration of their preferred light stimulus. ON and OFF transient cells have relatively low levels of spontaneous activity and spike only for a brief duration to their preferred stimulus.
By blocking glutathione peroxidase, which degrades low concentrations of free H2O2 to water, with MCS; (1 mm), the resulting increase in cytoplasmic H2O2 produced strikingly different effects in sustained and transient αRGCs stimulated by 1.2 s light steps, and the changes in spiking could generally be reversed on washout of MCS (Fig. 1). In ON transient cells (Fig. 1B), MCS addition significantly increased the maximum spike rate (142 ± 13% of control, p = 0.038, n = 8) and the number of spikes during the light response (121.7 ± 39% of control, p = 0.009), while in ON sustained RGCs (Fig. 1A) both metrics were significantly reduced (maximum spike rate: 51 ± 7% of control, p = 9.2 × 10−6; number of spikes: 53 ± 1% of control, p = 0.007, n = 8) with responses becoming much more transient in nature. This general trend was also observed in OFF sustained cells, albeit to a lesser degree, as only the number of spikes was reduced (Fig. 1C; 55 ± 1% of control p = 0.01, n = 9), and no effect was observed in OFF transient cells (Fig. 1D). Two more non-αRGC types were tested to evaluate the generality of this trend: M1 intrinsically photosensitive RGCs, with high levels of spontaneous activity and prolonged light responses; and unlabeled ON-OFF transient cells targeted using their large soma size. Light-evoked spiking in M1 cells was strongly reduced by MCS (Fig. 1E; spike rate: 15 ± 4% reduction, p = 1.7 × 10−5; spike number: 35 ± 9% reduction, p = 0.0006, n = 12), although in these cells recovery postwash was not observed. ON-OFF transient cells showed an increase in excitability during the ON component of the response (Fig. 1F; spike number: 340 ± 106% increase, p = 0.04, n = 5).
MCS revealed an interesting trend in its effects on αRGC excitability by elevating endogenous H2O2. In general, αRGCs with high metabolic demand produced fewer spikes after treatment, while cells with a relatively low metabolic demand responded with heightened excitability. However, this experimental paradigm does not eliminate the possibility that the effects of MCS were exerted on the presynaptic network. In fact, since bath-applied MCS likely exerts actions on all pre-RGC neuronal circuitry and glial function, by comparing the RGC firing pattern changes and rates in Figure 1 with those in Figure 2 (see also Fig. 4), which use ROS-increasing agents that act only on the recorded RGC, these data support the conclusions that major effects can be largely attributed to intrinsic RGC mechanisms.
Somatically applied H2O2 or glutathione modulate sustained but not transient RGC light responses in opposite directions
To avoid presynaptic effects, a relatively low (100 μm) concentration of H2O2 was applied using a puffer pipette located close to the cell soma. This allowed the rapid onset of effects and limited them largely to the cell soma. In ON and OFF sustained αRGCs, H2O2 caused a substantial reduction in light-evoked spiking (Fig. 2A: ON sustained cells: spike rate, 23.3 ± 11.3% reduction; p = 0.002; spike number, 8.7 ± 3.9% reduction; p = 0.003, n = 6; Fig. 2C: OFF sustained cells: spike number, 34.5 ± 11.0% reduction; p = 0.02; n = 7). The tendency for the light response in sustained αRGCs to become more transient in MCS also occurred following H2O2 application. In transient αRGCs, there were no significant changes following somatic H2O2 (Fig. 2B: ON transient cells, 25% reduction, n = 7; Fig. 2D: OFF transient cells, 5% reduction, n = 11).
The same protocol was used to apply a membrane-permeant analog of the antioxidant GME (5 mm). GME was puffed on the cell for ∼30 s, then the puffer pipette was removed and light responses were recorded following 10–15 min to allow the reduction of endogenous ROS. In sustained cells, GME increased light-evoked activity (Fig. 2A: ON sustained cells: spike number, 224 ± 25.8% increase; p = 0.006; n = 9; Fig. 2C: OFF sustained cells: spike number, 191 ± 31.9%; p = 0.01; n = 8). GME applied to transient RGCs caused insignificant increases (Fig. 2B: ON transient cells, n = 8; Fig. 2D: OFF transient cells, n = 8). These results suggest that endogenous ROS act as powerful intrinsic neuromodulators in RGCs having high metabolic rates but have essentially no effect in cells with much lower metabolic demand.
Compared with bath application, this method of limiting drug application to the vicinity of the cell soma reduces effects on large numbers of upstream neurons but does not rule out modest penetration of the drugs into presynaptic networks. To isolate stimulus and response to individual αRGCs, currents step of 1 s duration were injected to drive cell spiking directly. Spontaneous activity was eliminated by frequently adjusting each RGC resting potential to be near −72 ± 3 mV using steady-state injection of hyperpolarizing current (−50 to −200 pA). We fit and quantified the nonlinear, cumulative spiking response during injected current steps with a sigmoidal curve and report the half-maximal injected current (I1/2) and maximal spike rate of the fit. Sustained αRGCs showed significant reductions in spike number. In ON sustained cells, this was 23.4 ± 13.2% of control (p = 0.002, n = 6; Fig. 3A), and in OFF sustained cells the reduction was 43.4 ± 8.2% of control (p = 0.02, n = 7; Fig. 3C). There was no significant change to I1/2. The application of GME produced significant increases in spike number in OFF sustained cells of 141.6 ± 15.0% (p = 0.01, n = 8; Fig. 3C). GME also caused significant leftward shifts in I1/2 in sustained cells. ON sustained cells showed a shift of 35.4 ± 7.1 pA (p = 0.001, n = 9; Fig. 3A), while OFF sustained cells showed a shift of 24.1 ± 7.2 pA (p = 0.004, n = 8; Fig. 3C). Despite an apparent trend to follow sustained αRGCs, H2O2 and GME failed to significantly change parameters of the spike response to injected current steps in transient cells (Fig. 3B,D).
These results suggest that H2O2 dramatically modulates the excitability of sustained αRGCs but has little to no effect on transient cells. We noted that reducing endogenous ROS with a membrane-permeant analog of the endogenous antioxidant glutathione dramatically increased excitability in sustained cells, suggesting that endogenous the ROS level in sustained cells is high enough to modulate excitability in an ongoing manner. Increases in excitability occurred in sustained cells in response to both light and steps of current, suggesting that ROS modulates the intrinsic properties of sustained RGCs rather than having its actions via inputs from presynaptic networks.
Increasing mitochondrial ROS with intracellular application of the Complex III poison, antimycin-A, replicates the reduction in excitability in sustained cells
Although focal application of H2O2 and GME should largely limit the effects on presynaptic circuitry, the effects of altered ROS levels on the intrinsic properties of the αRGCs were further isolated by including oxidant-modulating agents in the recording pipette. We used catalase (200–500 IU/ml), a naturally present enzyme that speeds decomposition of H2O2 to H2O and O2 or antimycin-A (2 μm), which blocks Complex III (cytochrome reductase) of the electron transport chain, increasing mitochondrial superoxide. ATP was present in the internal solution, so this change cannot be attributed to reductions in energy supply. Like H2O2, antimycin-A caused significant reductions in light-evoked spiking in sustained αRGCs. The ON sustained cell spike rate was reduced to 63.2 ± 11.2% of control (same cell, baseline recordings following break-in; p = 0.03, n = 7), and the number of spikes was reduced to 36.3 ± 16.9% of control (p = 0.03, n = 7; Fig. 4A). The OFF sustained cell spike rate was reduced to 61.2 ± 7.7% of control (p = 0.01, n = 14), and spike number was reduced to 43.6 ± 9.9% of control (p = 0.0005, n = 14; Fig. 4C). Changes in spike rate and number mostly occurred within 10 min after break-in and remained stable over 30 min of recording. Recordings from ON (n = 5) and OFF (n = 5) sustained cells over 30 min with control internal solution (Fig. 4Av,vi,Cv,vi, black) did not show significant run-up or run-down, indicating that these changes are unlikely to be because of bleaching or time-dependent degradation in cell condition. When catalase was included in the recording pipette, effects opposite to those of antimycin-A occurred in sustained cells and were similar to GME effects. Both ON and OFF sustained cells underwent significant increases in excitability within 10 min post break-in and were stable for 30 min. ON sustained cells showed spike rates of 134.9 ± 19.3% of control (p = 0.001) and spike numbers at 171.2 ± 20.3% of control (p = 0.007, n = 9; Fig. 4A), while OFF sustained cells showed spike rates of 138.0 ± 26.0% of control (p = 0.006) and spike numbers of 236.9 ± 42.3% of control (p = 0.03, n = 7; Fig. 4C). In transient cells, antimycin-A generally increased the number of spikes produced during light responses, with ON transient cells showing spike numbers of 164.2 ± 22.1% of control (p = 0.009, n = 8; Fig. 4B) and OFF transient cells with spike numbers of 188.5 ± 53% of control (p = 0.02, n = 11; Fig. 4D). Like GME, catalase failed to significantly alter excitability in these cells.
The effects of antimycin-A and catalase were further assessed in sustained and transient αRGCs using injected current. Here we found that antimycin-A significantly reduced maximal spike rate in sustained cells (Fig. 5A: ON sustained cells: 33.9 ± 8.9% of control; p = 0.001; n = 8; Fig. 5C: OFF sustained cells: 75.6 ± 10% of control; p = 0.0007; n = 13) but not in I1/2. After 30 min of catalase, there were significant increases in maximal spike rate in sustained cells (Fig. 5A: ON cells: 124.4 ± 11.7% of control; p = 0.047; n = 7; Fig. 5C: OFF cells: 145.7 ± 14.9% of control; p = 0.02; n = 5). Catalase did not change the I1/2 significantly in sustained cells. Whereas there was generally an increase in the number of light-evoked spikes in transient cells (Fig. 4), there was no significant change in current-evoked spiking (Fig. 4), and this effect did not stabilize within 30 min. No significant change in transient cells with catalase was observed (Fig. 5B,D), suggesting that at baseline there is little to no modulation of transient cells by a standing level of ROS.
Biophysical mechanisms underlying the modulation of action potential efficiency and frequency by ROS-mediated feedback in RGCs
Given the changes in response to injected current in sustained cells we hypothesized that increased spiking activity of αRGCs drives elevations in intracellular ROS levels and that these elevated ROS levels modulate the gating of the ion channels underlying action potential generation. To assess how ROS affect the NaV channels that underlie the intrinsic excitability of each αRGC type and to assess biophysical differences between sustained and transient αRGCs, changes in the shape of action potentials in sustained versus transient cells following localized puff application of 100 μm H2O2 were determined. At baseline, transient cells produced relatively few spontaneous action potentials, so we injected Gaussian white noise current (μ = 0 pA2, σ = 16 pA2, 50 Hz sampling rate; MATLAB random number generator) to modestly increase spike generation (Fig. 6A–D). Somatically applied 100 μm H2O2 caused significant reductions in action potential height in all cell types (ON sustained: p < 1 × 10−10, n = 4 cells; OFF sustained: p < 1 × 10−10, n = 8 cells; OFF transient: p < 1 × 10−10, n = 5 cells; ON transient: p = 0.0001, n = 4 cells; Mann–Whitney U test) and increases in half-width in all except ON transient cells (ON sustained, p < 1 × 10−10; OFF sustained, p < 1 × 10−10; OFF transient, p = 0.01). However, the effect in sustained cells was much larger in all cases. H2O2 reduced spike amplitude in ON transient cells by ∼5% versus ∼45% in ON sustained cells, and half-width increased by ∼250 µs versus ∼1.7 ms in the same cells. In OFF cells, the same general pattern emerged with ∼5% reduction in spike height in OFF transient cells versus ∼25% in ON sustained cells and a 250 versus 350 µs increase in half-width. The slope of the membrane voltage during the action potential (dV/dt) was calculated as an approximation of NaV channel availability (Colbert et al., 1997; Weick and Demb, 2011). H2O2 caused a significant reduction in dV/dt in all cell types (ON sustained, p < 1 × 10−10; OFF sustained, p < 1 × 10−10; OFF transient, p < 1 × 10−10; ON transient cells p = 5.7 × 10−9); however, in sustained cells the reduction was much larger than in transient RGCs (Fig. 6A,C: ON cells: dV/dt ∼65% of control in sustained vs 85% of control in transient; Fig. 6B,D: OFF cells: ∼60% of control in sustained vs 95% of control in transient).
Observing the properties of action potentials suggested that the reduction in spiking in sustained αRGCs might be because of reductions in NaV current. To test this mechanism more explicitly, the nucleated patch technique was used to assess the effects of H2O2 on whole-cell NaV current in identified αRGCs. We identified αRGCs by extracellular light response (Fig. 7B–E), and then, following nucleation, 100 μm H2O2 was puff applied from a nearby pipette (Fig. 7A).
In nucleated patch recordings, inward currents could be completely blocked by 1 μm TTX (Fig. 7Ai,ii). The 100 μm H2O2 significantly reduced peak current in sustained but not transient αRGCs (Fig. 7F,G: sustained cells: 68.9 ± 2.7% of control; n = 15; p = 0.0002; transient cells: 106.1 ± 3.4% of control; n = 14). The inactivation curves for nucleated patches from αRGCs were left shifted with a mean inactivation half-maximal voltage (V1/2) of −102.4 ± 5.3 mV. Nucleated patch recordings from other RGC types (e.g., M1 RGCs) did not present with this shifted inactivation V1/2 (−66.7 ± 1.3 mV), suggesting that this was not a limitation of the technique itself but was caused by some other feature of αRGCs, possibly their unusually large somata. This limited our ability to assess the voltage dependence of inactivation as a possible mechanism underlying the reduced current following H2O2 (Fig. 7I). However, there were significant leftward shifts in activation in both sustained and transient cells (Fig. 7H; sustained cells: −3.5 ± 0.9 mV; p = 0.0017; transient cells: −3.6 ± 1.4 mV; p = 0.02).
We noted that elevating ROS typically made the spiking response to either current injection or light to become more transient in sustained αRGCs and the nucleated patch recordings implicated reduction of NaV current as a plausible mechanism. This is similar to studies of NaV1.1 knock-out mice, which show a similar deficit in repetitive spiking (Yu et al., 2006; Ogiwara et al., 2007). In OFF transient αRGCs, a similar increase in transience was observed in response to injected current using 500 nm 4,9-anhydrotetrodotoxin (4,9ahTTX; Werginz et al., 2020). At this concentration, 4,9ahTTX blocks both NaV1.1 and NaV1.6 (Griffith et al., 2019; Denomme et al., 2020; Werginz et al., 2020). However, a highly specific NaV1.1/1.3 channel blocker, ICA121431 (Griffith et al., 2019), is selective for NaV1.1 and 1.3 over the NaV1.2 and NaV1.6 channel isoforms also expressed by RGCs (Van Hook et al., 2019). Although ICA121431 is nonselective between NaV1.1 and NaV1.3, NaV1.3 is normally expressed primarily in the developing, nonadult nervous system (Waxman et al., 1994; Felts et al., 1997) and is also upregulated after injury (Waxman et al., 1994; Gastaldi et al., 1997). Because of these constraints, we consider the most likely NaV channel isoform blocked by ICA121431 in this current study to be NaV1.1, notwithstanding earlier detection of NaV1.3 mRNA in RGC somata using in situ hybridization (Fjell et al., 1997). Regardless of the isoform responsible, blocking NaV1.1/1.3 channels replicates the selective effects of ROS elevation on sustained cells and prevents additional ROS effects.
Somatic application of ICA121431 (500 nm) was used to first assess whether blocking NaV1.1/1.3 channels might have different effects on sustained and transient αRGCs (Fig. 8). ICA121431 caused the light response of sustained αRGCs to become much more transient, mimicking the effect of H2O2, antimycin-A, and MCS (Fig. 8A: ON sustained cells: spike number, 38.3 ± 5.1% of control; p = 0.0002; n = 6; Fig. 8C: OFF sustained cells: spike number, 31.5 ± 6.3% of control; p = 0.01; n = 9). Maximal spike rate tended to be reduced, but this was not significant. In transient αRGCs, the effect of ICA121431 was much less pronounced, and, although a mean reduction in the number of spikes was seen in OFF transient αRGCs (n = 5), this did not reach significance (Fig. 8B,D). Following the application of ICA121431, the puffer pipette was replaced with one containing both ICA121431 and 100 μm H2O2. When NaV1.1/1.3 was blocked with ICA121431, there was little additional effect of H2O2 in either sustained or transient αRGCs. The one exception was that, whereas ICA121431 alone did not significantly reduce maximal spike rate in ON sustained αRGCs, the combination of ICA121431 plus H2O2 did (p = 0.01; one-way ANOVA with post hoc Tukey's HSD test). The treatments with ICA121431 and ICA121431 plus H2O2 are not significantly different from each other (p = 0.68).
The effects of ICA121431 alone and in combination with H2O2 on the response to injected current was also tested. ICA121431 caused a significant reduction in maximal spike rate in sustained αRGCs (Fig. 9A: ON sustained cells: spike rate, 60.0 ± 10.4% of control; p = 0.04; n = 6; Fig. 9C: OFF sustained cells: spike number, 49.7 ± 12.4% of control; p = 0.002; n = 5). Similar to the light response, the response to injected current became much more transient following treatment with ICA121431. When H2O2 was combined with ICA121431, we did not note any further reduction in spike rate, and in fact, there was a slight but insignificant increase in maximal spike rate. I1/2 was not changed in sustained αRGCs in either treatment, and significant changes to the response to injected current in ON and OFF transient αRGCs in response to either treatment were not observed (Fig. 9B,D).
Changes to the shape of action potentials driven by either 500 nm ICA121431 alone or in combination with 100 μm H2O2 were also investigated. In earlier experiments, as per Figure 6, the slope of the membrane voltage during the action potential (dV/dt) was most indicative of changes in NaV channel availability, so we focused specifically on this metric comparing the actions of ICA121431 alone or in combination with H2O2 on spiking in Figure 10. We found that dV/dt was significantly reduced by ICA121431 in both types of sustained αRGCs (ON sustained cells: ∼60% of control; p < 1 × 10−10; OFF sustained cells: ∼65% of control; p < 1 × 10−10; Kruskal–Wallis test with post hoc Dunn's test for multiple comparisons) and further treatment of 100 μm H2O2 + 500 nm ICA121431 caused a small additional decrease in both cell types (ON sustained cells, ∼55% of control: p < 1 × 10−10 vs control and ICA121431 alone; OFF sustained cells, 50% of control: p < 1 × 10−10 vs control and ICA121431 alone). In contrast to sustained αRGCs, relatively small reductions in transient αRGCs were noted (ON transient cells: ICA121431, ∼90% of control; ICA121431 + H2O2, ∼90% of control; p = 0.63 vs control; p = 0.005 vs ICA121431 alone; OFF transient cells: ICA121431, 95% of control, p = 0.46; ICA121431+ H2O2, ∼90% of control, p = 0.0024). These results demonstrate that the majority of the effects observed with H2O2, and presumably with other ROS-elevating treatments, can be replicated by blocking NaV1.1/1.3 in sustained αRGCs.
Discussion
Neurons have highly diverse electrogenic phenotypes. Understanding why some biophysical phenotypes are favored is a key step in deepening our understanding of what neurons are optimized for in diverse neural tissues. Minimizing energetic cost is emerging as a powerful evolutionary constraint on successful neural phenotypes. We investigated whether ROS, produced as an inevitable consequence of ATP production, might act intrinsically to modulate the excitability of sustained RGCs having high metabolic demand. A wide variety of ROS-manipulating reagents revealed that in sustained but not transient RGCs, ROS elevation reduced excitability. Reducing ROS by augmenting the natural antioxidant abilities of the cells increased excitability in sustained RGCs with no effect on transient RGCs. The expression of NaV1.1/1.3 channels, found to be necessary for sustained αRGCs to produce tonic spiking in response to prolonged light, also endows these cells with a feedback mechanism that limits spike rate. This connection between electrogenic phenotype and energy-conserving adaptation is broadly consistent with metabolically efficient coding, although further work is necessary to test the relationship between information bandwidth and energetic demand with and without NaV1.1/1.3 function.
Sustained, but not transient, αRGCs have tonically elevated ROS levels that intrinsically modulate excitability
Neurons are constantly producing ROS and superoxide as a consequence of ATP production. Tonically elevated ROS is known to cause neurodegeneration, so neurons typically use antioxidants (e.g., catalase, glutathione peroxidase, superoxide dismutase, vitamins C and E) to control ROS levels. Here we provide evidence that ROS levels act as neuromodulators. Augmenting the natural antioxidant capacity of RGCs with catalase or GME increased excitability dramatically in sustained, but not transient, αRGCs. Both light-evoked and current-evoked spiking increased in sustained αRGCs following ROS reduction, suggesting that the mechanism of action was on cellular intrinsic properties rather than synaptic input. This implies that tonically elevated ROS levels act as a feedback mechanism in sustained αRGCs to reduce excitability, and that excitability in these cells can be manipulated by altering the production of antioxidants.
In contrast, transient αRGCs were essentially unaffected by reducing standing ROS levels. This suggests that either ROS levels are low enough under normal circumstances that further reduction is inconsequential or the mechanism of action leading to ROS-dependent reductions is lacking in transient αRGCs. To test how ROS affect RGC function, ROS levels were modulated with three different approaches. The results of ROS elevation were consistent in sustained αRGCs. MCS, antimycin-A, and somatically applied H2O2 reduced excitability in response to light and injected currents, making responses more transient. The similarity between sustained αRGCs following ROS elevation and transient αRGCs under control conditions suggested that ROS might act on the intrinsic properties that allow sustained RGCs to produce tonic spiking in response to light and current injection. The effects on transient αRGCs were more dependent on the mechanism used to elevate ROS and ranged from no effect to increased excitability in response to light, with little or no effect on injected current.
ROS reduce excitability in sustained αRGCs by reducing NaV1.1/1.3 channel currents
The results implicate NaV channels in having a principal role in the mechanism of action by which ROS mediate reductions in excitability in sustained αRGCs. First, it was found that the height, half-width, and temporal derivative (i.e., dV/dt) of the action potentials were strongly altered in sustained cells, but not in transient αRGCs. In particular, the slope of the rising phase of the action potential is established as a solid approximation of NaV current availability in neurons (Colbert et al., 1997; Weick and Demb, 2011). Compared with transient αRGCs, dV/dt was greatly reduced by somatic application of H2O2 in sustained cells. We isolated nucleated patches from αRGCs identified by their extracellular light responses and found significant reductions in peak NaV current following H2O2 in sustained but not transient cells. Surprisingly, in αRGCs, but not in other RGC types from which nucleated patches were pulled, the inactivation V1/2 was very left shifted, and a hyperpolarizing prepulse close to −120 mV was required for full NaV current activation, precluding meaningful analysis of the underlying voltage-dependent gating kinetics.
Our observations suggest that NaV channels play a significant role in RGC modulation by ROS. RGCs express NaV1.1, 1.2, 1.6, 1.8, and perhaps 1.3 (Fjell et al., 1997; Boiko et al., 2003; Van Wart and Matthews, 2006; Mojumder et al., 2007; O'Brien et al., 2008; Smith et al., 2017). While both NaV1.1 and 1.6 have been implicated in sustained repetitive spiking, ICA121431, a blocker of NaV1.1 (and NaV1.3) is highly selective, while the common blocker of NaV1.6 (4,9 anhydro-tetrodotoxin) is not selective between NaV1.1 and 1.6 (Griffith et al., 2019; Denomme et al., 2020). When NaV1.1/1.3 channels in sustained αRGCs were blocked with somatic application of ICA121431, their light responses were rendered more transient, similar to the effects of ROS elevation. Further application of a mix of ICA121431 and 100 μm H2O2 in sustained αRGCs had only minor additive effects on light-evoked and current-evoked spiking, suggesting actions on the same target. This possibility is supported as neither ICA121431 alone or in combination with H2O2 caused significant reductions in transient cells. When the effects of ICA121431 alone and in combination with H2O2 on dV/dt were assessed, we found that in sustained cells, almost the entire effect of H2O2 on dV/dt could be eliminated following treatment with ICA121431.
The results with the selective blocker ICA121431 are consistent with robust NaV1.1/1.3 channel function in sustained αRGCs, not transient αRGCs. This would facilitate the high-frequency repetitive spiking and high spontaneous activity observed in these cells and create a built-in adaptive mechanism reducing firing rate when ROS levels rise and increasing excitability when ROS levels fall. These results imply that transient cells are much less affected by ROS because of their having fewer redox-sensitive NaV channel isoforms. Previous reports provide evidence of redox modulation of NaV channels (Evans and Bielefeldt, 2000; Fukuda et al., 2005; Kassmann et al., 2008) and differences in oxidative modulation between molecular isoforms (Kassmann et al., 2008; Schlüter and Leffler, 2016).
Transient RGCs increase in excitability following ROS elevation
Mitochondrial dysfunction drives optic neuropathies because of genetic mutations (e.g., Leber hereditary optic neuropathy, Leigh disease, dominant optic atrophy) and environmental toxins (Pilz et al., 2017). In most cases, including glaucoma, ROS accumulation is a key feature of neurodegeneration (Tezel, 2006). In optic nerve crush or transection models, where RGC ROS levels are elevated (Kanamori et al., 2010; Sayır et al., 2013; Fan et al., 2017), transient cells degenerate more rapidly (Della Santina et al., 2013; Tran et al., 2019). Our results using the Complex III toxin antimycin-A or MCS, which increase ROS, showed increased excitability in both ON and OFF transient αRGCs over 30 min. Unlike the actions of antimycin-A in sustained cells that essentially stabilized after 10 min, in transient cells the effects of antimycin-A were slower in onset and did not stabilize during the 30 min of recording time.
The mechanisms underlying changes in transient αRGC excitability bear further investigation given their relevance to chronically elevated ROS levels in disease states. It is known that H2O2 diffuses through membranes and can act presynaptically, increasing glutamate release in ventral dorsal horn neurons (Ohashi et al., 2016a). Our direct application of H2O2 to RGC somata produced no significant effect in transient αRGCs, nor did antimycin-A produce significant changes to responses to injected current, suggesting that the changes involved upstream or synaptic inputs. Earlier studies in isolated RGCs (Aizenman et al., 1989) showed that NMDA receptors were potentiated under reducing conditions and that there was no effect on AMPA/kainate receptors, suggesting that a direct action of ROS-potentiating glutamate receptors in transient RGCs was unlikely.
Inhibitory GABA and glycine synaptic currents in RGCs are decreased by oxidants and potentiated by antioxidants (Pan et al., 1995). In contrast, ROS increase some GABAergic currents (Accardi et al., 2014; Penna et al., 2014; Ohashi et al., 2016a, b). This suggests additional mechanisms by which ROS might both inhibit and excite αRGCs during chronically elevated ROS levels leading to pathologic hyperexcitability.
Our findings suggest that, compared with sustained αRGCs, transient αRGCs have much lower levels of NaV1.1/1.3 expression (judged by their minor ICA121431 block) and that while this contributes to their reduced basal metabolic demand, it is consistent with the lack of the pronounced ROS-mediated inhibitory feedback mechanism shown here in sustained αRGCs. During antimycin-A-induced ROS increases, transient RGCs became more excitable. It will be important to pinpoint the exact mechanism underlying this increased excitability and determine its role in neurodegenerative states.
In conclusion, these results provide compelling evidence that endogenous ROS act as modulators in sustained, not transient, αRGCs, and that the mechanisms underlying this action are mediated by the intrinsic properties of the RGCs rather than arising from presynaptic mechanisms. Although substantial literature indicates that ROS modify the excitability of neurons, these novel findings specifically link intrinsic metabolic demand to a significant neuromodulatory role for ROS within neurons, here RGCs, in which the cell subtypes have widely varying basal metabolic rates and intrinsic biophysical properties.
Footnotes
This work was supported by National Institutes of Health (NIH) Grant F32-EY-032401 (B.J.S.), the Glaucoma Research Foundation (S.B.), the Plum Foundation (S.B.), the Keck Foundation (S.B.), NIH Grant R01-EY-04067 (N.C.B.), Veterans Administration Senior Career Scientist Award (N.C.B.), and an Unrestricted Grant from Research to Prevent Blindness, Inc. (to the UCLA Department of Ophthalmology). We thank Dr. Anna Matynia for providing OPN4EGFP mice and Dr. Rikard Frederiksen for assisting with light calibration.
The authors declare no competing financial interests.
- Correspondence should be addressed to Steven Barnes at sbarnes{at}doheny.org